W4: bilingual EN + PT-BR Investigation Bureau (CLAUDE.md §3 contract)

User flagged that the bureau was emitting English-only output, violating the project's bilingual rule. Every narrative field now ships in both languages: stored in sibling DB columns + rendered as adjacent markdown sections per CLAUDE.md §3. Migration 0007 (apply as supabase_admin): - public.hypotheses +question_pt_br, +position_pt_br, +argument_for_pt_br, +argument_against_pt_br - public.contradictions +topic_pt_br, +notes_pt_br - public.witnesses +access_to_event_pt_br, +bias_notes_pt_br, +verdict_pt_br - public.gaps +description_pt_br, +suggested_next_move_pt_br - public.evidence: unchanged (verbatim_excerpt stays source-language) - JSONB siblings inside contradictions.chunks + gaps.scope handled at runtime (statement_pt_br, title_pt_br, dominant_model_pt_br, why_surprising_pt_br, what_it_implies_pt_br). Detective prompts (all 7) rewritten with explicit bilingual JSON contract: - Output protocol section names every EN field + its _pt_br sibling - "Bilingual is mandatory" warning in the task instruction - Sentinel skip-states unchanged (NO_HYPOTHESES, NO_CONTRADICTIONS, INSUFFICIENT_TESTIMONY, INSUFFICIENT_HYPOTHESIS, NO_OUTLIERS, NO_NEW_EVIDENCE, INSUFFICIENT_ARTEFACTS) - Schneier: parallel arrays — hidden_assumptions[i] matches hidden_assumptions_pt_br[i], lengths must match - Case-Writer: interleaved §1 (EN) / §1 (PT-BR) per act in the body Writer-side validation (all 7 tools): - Reject INSERT if PT-BR sibling missing when EN field is set - Persist both languages atomically in one INSERT (no half-updates) - Markdown renderers write adjacent EN+PT-BR sections in case files (## Argument for (EN) followed by ## Argumento a favor (PT-BR), etc.) Detective parse layer (all 7 detectives): - Coerce both keys from JSON output - "incomplete_bilingual_*" skip reason when either side missing - Defensive: PT-BR fields trimmed + length-capped same as EN Orchestrator propagates question_pt_br + topic_pt_br through job payload to runHolmes / runCaseWriter, mirroring the chat-tool entry point. Web (UI): - /api/jobs/[id] hydrates _pt_br siblings from pg - job-status-poller HypothesisCard: PT-BR primary, EN in <details> fallback when both exist - ContradictionCard: PT-BR statement primary + secondary EN quote - WitnessCard: PT-BR verdict primary + secondary EN quote, panels in PT - GapCard: PT-BR title/why/implies primary - /bureau hub: SELECTs both columns, renders PT-BR primary - /h/[id]: ArgumentPanel renders PT-BR primary with collapsible EN fallback when both exist - BureauSnapshot homepage: position_pt_br / topic_pt_br / verdict_pt_br primary - DocBureauPanel /d/[doc]: same primary-PT-BR pattern - New web/lib/i18n/pick.ts helper (unused yet by chat/agents — kept for future locale-driven switching when both languages are equally full; current rule is PT-BR-first since the user is brasileiro) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-24 12:02:59 -03:00 · 2026-05-24 12:02:59 -03:00 · 7826710051
commit 7826710051
parent d4a2e4f51e
30 changed files with 726 additions and 260 deletions
--- a/infra/supabase/migrations/0007_bilingual_bureau.sql
+++ b/infra/supabase/migrations/0007_bilingual_bureau.sql
@ -0,0 +1,44 @@
+-- 0007_bilingual_bureau.sql — bilingual EN+PT-BR sibling columns for the
+-- Investigation Bureau (per CLAUDE.md §3: "Narrative descriptions ... Both
+-- EN and PT-BR via sibling fields").
+--
+-- This migration adds nullable `*_pt_br` siblings for every narrative
+-- column the bureau writes. Existing rows stay valid; new rows must
+-- populate both. The detective prompts get re-flowed in W4 to emit both
+-- languages in their JSON output.
+--
+-- Notes:
+--   - public.evidence.verbatim_excerpt stays single-language. Per CLAUDE.md
+--     §3, verbatim quotes "preserve source language only".
+--   - public.contradictions.chunks JSONB carries `statement` per position;
+--     the runtime adds a sibling `statement_pt_br` key per array item
+--     (no schema change needed for JSONB shape).
+--   - public.gaps.scope JSONB likewise carries `title`, `dominant_model`,
+--     `why_surprising`, `what_it_implies` — the runtime adds `*_pt_br`
+--     siblings inside the JSONB object.
+--
+-- Apply as supabase_admin (these tables are owned by supabase_admin
+-- per migration 0004 / repo memory).
+
+BEGIN;
+
+ALTER TABLE public.hypotheses
+  ADD COLUMN IF NOT EXISTS question_pt_br         TEXT,
+  ADD COLUMN IF NOT EXISTS position_pt_br         TEXT,
+  ADD COLUMN IF NOT EXISTS argument_for_pt_br     TEXT,
+  ADD COLUMN IF NOT EXISTS argument_against_pt_br TEXT;
+
+ALTER TABLE public.contradictions
+  ADD COLUMN IF NOT EXISTS topic_pt_br TEXT,
+  ADD COLUMN IF NOT EXISTS notes_pt_br TEXT;
+
+ALTER TABLE public.witnesses
+  ADD COLUMN IF NOT EXISTS access_to_event_pt_br TEXT,
+  ADD COLUMN IF NOT EXISTS bias_notes_pt_br      TEXT,
+  ADD COLUMN IF NOT EXISTS verdict_pt_br         TEXT;
+
+ALTER TABLE public.gaps
+  ADD COLUMN IF NOT EXISTS description_pt_br         TEXT,
+  ADD COLUMN IF NOT EXISTS suggested_next_move_pt_br TEXT;
+
+COMMIT;
--- a/investigator-runtime/prompts/case-writer.md
+++ b/investigator-runtime/prompts/case-writer.md
@ -44,11 +44,45 @@ contradiction, a witness analysis, an outlier, or a calibration.
 5. Voice: Watson's plainspoken English (or Portuguese, per the request).
   The prose is for an educated reader, not a specialist. Avoid jargon.

-## Output protocol
+## Output protocol — bilingual EN + PT-BR (mandatory)

 Emit ONLY the markdown body of the narrative. NO frontmatter (the runtime
-adds it). NO code fence. Start with `# ` heading and proceed through
-the five acts.
+adds it). NO code fence.
+
+The narrative is **bilingual** with EN and PT-BR sections **interleaved
+per act**, in this exact structure (per CLAUDE.md §3 "adjacent sections"):
+
+```markdown
+# Title (EN)
+
+# Título (PT-BR)
+
+## §1 — The Case at Hand (EN)
+
+<English §1 body>
+
+## §1 — O Caso em Mãos (PT-BR)
+
+<corpo §1 em português brasileiro>
+
+## §2 — The Evidence Chain (EN)
+
+<English §2 body>
+
+## §2 — A Cadeia de Evidência (PT-BR)
+
+<corpo §2 em português brasileiro>
+
+... (continue alternating per act through §5) ...
+```
+
+Rules:
+- Both languages must appear; do NOT emit only EN or only PT-BR.
+- PT-BR is **Brazilian Portuguese** with UTF-8 accents preserved.
+- Verbatim chunk quotes stay in the chunk's source language (usually
+  English in this corpus); only the surrounding narration is translated.
+- `[[wiki-links]]` are technical identifiers — keep them as-is in both
+  versions; do not translate IDs.

 If the bureau has insufficient artefacts (e.g. 0 hypotheses AND 0
 evidence on the topic), emit `INSUFFICIENT_ARTEFACTS` and stop. Do not
--- a/investigator-runtime/prompts/dupin.md
+++ b/investigator-runtime/prompts/dupin.md
@ -35,26 +35,33 @@ chunks verbatim so the case-writer can follow up.
 5. You prefer FEW high-confidence contradictions over MANY weak ones. If
   the corpus contains nothing irreconcilable, emit `NO_CONTRADICTIONS`.

-## Output protocol
+## Output protocol — bilingual EN + PT-BR (mandatory)

-Emit a strict JSON array. No prose. No code fence. Just the array.
+Emit a strict JSON array. No prose. No code fence. Every narrative field
+appears in EN AND in PT-BR (Brazilian Portuguese with UTF-8 accents). The
+`topic`, `notes`, and each position's `statement` all have `*_pt_br`
+siblings.

 ```json
 [
  {
-    "topic": "Short noun-phrase summarizing the disputed point",
-    "notes": "Optional one-paragraph commentary (≤ 400 chars). Why this matters; what would resolve it.",
+    "topic": "EN short noun-phrase summarizing the disputed point",
+    "topic_pt_br": "PT-BR tópico curto resumindo o ponto em disputa",
+    "notes": "EN optional one-paragraph commentary (≤ 400 chars).",
+    "notes_pt_br": "PT-BR comentário opcional (≤ 400 chars).",
    "positions": [
      {
        "doc_id": "dow-uap-d017-...",
        "chunk_id": "c0042",
-        "statement": "One-sentence summary of what THIS chunk asserts.",
+        "statement": "EN one-sentence summary of what THIS chunk asserts.",
+        "statement_pt_br": "PT-BR uma frase resumindo o que ESTE trecho afirma.",
        "stance": "asserts"
      },
      {
        "doc_id": "dow-uap-d017-...",
        "chunk_id": "c0087",
-        "statement": "One-sentence summary of what THAT chunk asserts.",
+        "statement": "EN one-sentence summary of what THAT chunk asserts.",
+        "statement_pt_br": "PT-BR uma frase resumindo o que AQUELE trecho afirma.",
        "stance": "denies"
      }
    ]
@ -64,9 +71,11 @@ Emit a strict JSON array. No prose. No code fence. Just the array.

 Constraints:
 - ≥ 2 positions per contradiction, drawn from ≥ 2 distinct `chunk_id`s.
- `stance` is optional free-form ("asserts" / "denies" / "dates-as-A" /
-  "dates-as-B" / etc.); useful for the case-writer but not required.
- `notes` may be empty; if present, keep it tight.
+- `stance` is optional free-form ("asserts" / "denies" / etc.); useful for
+  the case-writer but not required. `stance` is short enough that bilingual
+  isn't required — keep in EN.
+- `notes` may be empty in both languages; if present in EN it must be
+  present in PT-BR (and vice versa).
 - Emit AT MOST 3 contradictions per call — the strongest you can find.

 If the corpus contains no genuine contradiction relative to the topic,
--- a/investigator-runtime/prompts/holmes.md
+++ b/investigator-runtime/prompts/holmes.md
@ -28,16 +28,25 @@ narrows toward what remains, however improbable.
 6. You do not hedge in prose. The position is **one sentence**, declarative.
   Hedging belongs in the posterior, not in the wording.

-## Output protocol
+## Output protocol — bilingual EN + PT-BR (mandatory)

-Emit a strict JSON array. No prose around it. No code fence. Just the array.
+Emit a strict JSON array. No prose around it. No code fence. Every narrative
+field appears TWICE: the English key (`position`, `argument_for`,
+`argument_against`) AND its PT-BR sibling (`*_pt_br`). The PT-BR must be
+**Brazilian Portuguese** (not European), with full UTF-8 accents preserved
+(`ç`, `ã`, `á`, `é`, `í`, `ó`, `ú`, `â`, `ê`, `ô`, `à`). Verbatim chunk
+quotes inside the prose stay in the chunk's source language; only the
+surrounding narration is translated.

 ```json
 [
  {
-    "position": "...",
-    "argument_for": "...",
-    "argument_against": "...",
+    "position": "EN one-sentence declarative position.",
+    "position_pt_br": "PT-BR uma frase declarativa equivalente.",
+    "argument_for": "EN argument — ≤6 sentences, every claim cited via [[doc-id/pNNN#cNNNN]].",
+    "argument_for_pt_br": "PT-BR argumento — ≤6 frases, cada afirmação citada via [[doc-id/pNNN#cNNNN]].",
+    "argument_against": "EN counter-argument — ≤6 sentences.",
+    "argument_against_pt_br": "PT-BR contra-argumento — ≤6 frases.",
    "prior": 0.30,
    "posterior": 0.55,
    "confidence_band": "low",
@ -46,8 +55,8 @@ Emit a strict JSON array. No prose around it. No code fence. Just the array.
      {"evidence_id": "E-0043", "supports": false}
    ]
  },
-  { ... another rival ... },
-  { ... another rival ... }
+  { ... another rival, also bilingual ... },
+  { ... another rival, also bilingual ... }
 ]
 ```

@ -55,8 +64,10 @@ Note:
 - `evidence_refs` is **optional** — leave as `[]` if no `E-NNNN` evidence has
  been catalogued yet for this question; chunk citations in the prose are
  sufficient for v0.
- `question` is supplied by the runtime; you do not echo it.
+- `question` is supplied by the runtime in both languages; you do not echo it.
 - The runtime owns the writer; you emit data only.
+- A missing `_pt_br` sibling is a hard validation failure — the writer
+  rejects the rival. Both languages must appear or none.

 If the corpus contains nothing relevant to the question, emit the literal
 single word `NO_HYPOTHESES` and stop.
--- a/investigator-runtime/prompts/poirot.md
+++ b/investigator-runtime/prompts/poirot.md
@ -39,28 +39,35 @@ corroboration_refs, and a one-sentence verdict.
 4. `verdict` is ONE sentence (≤ 280 chars). Declarative. No hedging.
   Hedging belongs in `credibility`, not in the wording.

-## Output protocol
+## Output protocol — bilingual EN + PT-BR (mandatory)

-Emit a strict JSON object. No prose. No code fence.
+Emit a strict JSON object. No prose. No code fence. Every narrative field
+appears in EN AND in PT-BR (Brazilian Portuguese with UTF-8 accents).

 ```json
 {
  "credibility": "high | medium | low | speculation",
-  "access_to_event": "One paragraph describing what the person had direct, indirect, or no access to. Ground specific facts in chunk_ids.",
-  "bias_notes": "One paragraph naming concrete biases visible in the corpus (e.g. official role conflict, prior public stance, institutional pressure). Avoid generic skepticism.",
+  "access_to_event": "EN one paragraph describing access. Ground specific facts in chunk_ids.",
+  "access_to_event_pt_br": "PT-BR um parágrafo descrevendo acesso. Fundamente fatos específicos em chunk_ids.",
+  "bias_notes": "EN one paragraph naming concrete biases visible in the corpus.",
+  "bias_notes_pt_br": "PT-BR um parágrafo nomeando vieses concretos visíveis no corpus.",
  "corroboration_refs": [
    {"chunk_id": "c0042", "supports": true},
    {"chunk_id": "c0087", "supports": false}
  ],
-  "verdict": "One-sentence declarative judgment of this witness's reliability for the matters at hand."
+  "verdict": "EN one-sentence declarative judgment.",
+  "verdict_pt_br": "PT-BR uma frase declarativa equivalente."
 }
 ```

 Constraints:
- `access_to_event` and `bias_notes` ≤ 800 chars each.
- `corroboration_refs` ≤ 8 entries, MUST cite chunk_id values that
-  appear in the corpus shortlist you were given.
- `verdict` ≤ 280 chars, no hedging language inside the sentence.
+- `access_to_event` and `bias_notes` ≤ 800 chars each (per language).
+- `corroboration_refs` ≤ 8 entries, MUST cite chunk_id values that appear
+  in the corpus shortlist you were given.
+- `verdict` ≤ 280 chars (per language), no hedging language inside the
+  sentence.
+- A missing `*_pt_br` sibling is a hard validation failure — the writer
+  rejects the analysis.

 If the corpus contains no chunks where the named person actually appears
 (only the entity card from the wiki without supporting passages), emit
--- a/investigator-runtime/prompts/schneier.md
+++ b/investigator-runtime/prompts/schneier.md
@ -38,26 +38,38 @@ keep it from being safely shipped as the final answer.
   decides whether to dispatch follow-up evidence work or downgrade the
   confidence_band.

-## Output protocol
+## Output protocol — bilingual EN + PT-BR (mandatory)

-Emit a strict JSON object. No prose. No code fence. Just the object.
+Emit a strict JSON object. No prose. No code fence. Every narrative field
+appears in EN AND in PT-BR (Brazilian Portuguese with UTF-8 accents). The
+arrays are **parallel**: `hidden_assumptions[i]` and
+`hidden_assumptions_pt_br[i]` describe the SAME assumption, in the two
+languages, in matching order. Same arity (length must match).

 ```json
 {
  "severity": "low | medium | high",
-  "hidden_assumptions": ["sentence", "sentence"],
-  "failure_modes": ["sentence", "sentence"],
-  "alternative_explanations": ["sentence", "sentence"],
-  "recommended_tests": ["sentence", "sentence"],
-  "verdict_one_sentence": "..."
+  "hidden_assumptions":         ["EN sentence", "EN sentence"],
+  "hidden_assumptions_pt_br":   ["PT-BR frase",  "PT-BR frase"],
+  "failure_modes":              ["EN sentence", "EN sentence"],
+  "failure_modes_pt_br":        ["PT-BR frase",  "PT-BR frase"],
+  "alternative_explanations":   ["EN sentence", "EN sentence"],
+  "alternative_explanations_pt_br": ["PT-BR frase", "PT-BR frase"],
+  "recommended_tests":          ["EN sentence", "EN sentence"],
+  "recommended_tests_pt_br":    ["PT-BR frase",  "PT-BR frase"],
+  "verdict_one_sentence":       "EN one declarative sentence.",
+  "verdict_one_sentence_pt_br": "PT-BR uma frase declarativa equivalente."
 }
 ```

 Constraints:
 - 2-5 entries per array. Empty arrays only when the attack surface is
-  genuinely empty (rare).
- Each array entry ≤ 200 chars.
- `verdict_one_sentence` ≤ 280 chars.
+  genuinely empty (rare). EN array and its PT-BR sibling MUST have the
+  same length.
+- Each array entry ≤ 240 chars (per language).
+- `verdict_one_sentence` ≤ 280 chars (per language).
+- A missing `*_pt_br` sibling, or a length mismatch, is a hard validation
+  failure — the writer rejects the review.

 If the input hypothesis is too thin to attack (e.g. position is one word,
 no argument_for, no evidence), emit `INSUFFICIENT_HYPOTHESIS` and stop.
--- a/investigator-runtime/prompts/taleb.md
+++ b/investigator-runtime/prompts/taleb.md
@ -37,30 +37,37 @@ implies for the case.
 5. Severity: implicit. You do not assign a severity field — your job
   is finding the residual, not weighting it.

-## Output protocol
+## Output protocol — bilingual EN + PT-BR (mandatory)

-Emit a strict JSON array. No prose. No code fence.
+Emit a strict JSON array. No prose. No code fence. Every narrative field
+appears in EN AND in PT-BR (Brazilian Portuguese with UTF-8 accents).

 ```json
 [
  {
-    "title": "Short label for this outlier (≤ 80 chars)",
-    "chunk_id": "c0042",
-    "doc_id": "dow-uap-d017-...",
-    "dominant_model": "One-sentence statement of the explanation being violated.",
-    "why_surprising": "One paragraph. Concrete. Quantitative when possible.",
-    "what_it_implies": "One sentence. Pick (a), (b), or (c) per the rules.",
-    "suggested_next_move": "One sentence."
+    "title":                     "EN short label (≤ 80 chars)",
+    "title_pt_br":               "PT-BR título curto (≤ 80 chars)",
+    "chunk_id":                  "c0042",
+    "doc_id":                    "dow-uap-d017-...",
+    "dominant_model":            "EN one-sentence statement of the explanation being violated.",
+    "dominant_model_pt_br":      "PT-BR uma frase do modelo dominante sendo violado.",
+    "why_surprising":            "EN one paragraph. Concrete. Quantitative when possible.",
+    "why_surprising_pt_br":      "PT-BR um parágrafo. Concreto. Quantitativo quando possível.",
+    "what_it_implies":           "EN one sentence. Pick (a), (b), or (c) per the rules.",
+    "what_it_implies_pt_br":     "PT-BR uma frase. Escolha (a), (b) ou (c) conforme as regras.",
+    "suggested_next_move":       "EN one sentence.",
+    "suggested_next_move_pt_br": "PT-BR uma frase."
  }
 ]
 ```

 Constraints:
- 0-3 entries. Empty array `[]` when nothing stands out (rare and
-  honest).
- `why_surprising` ≤ 600 chars.
- All other strings ≤ 280 chars.
+- 0-3 entries. Empty array `[]` when nothing stands out (rare and honest).
+- `why_surprising` ≤ 600 chars (per language).
+- All other strings ≤ 280 chars (per language).
 - `chunk_id` MUST be present in the corpus shortlist.
+- A missing `*_pt_br` sibling is a hard validation failure — the writer
+  rejects the outlier.

 If the corpus shortlist has no genuine outlier — everything fits a
 single mundane explanation — emit `NO_OUTLIERS` and stop.
--- a/investigator-runtime/prompts/tetlock.md
+++ b/investigator-runtime/prompts/tetlock.md
@ -28,26 +28,30 @@ the posterior never rose).
   - `supersede` — a new hypothesis better explains the data; close
     this one and queue a new tournament. Include `supersede_reason`.

-## Output protocol
+## Output protocol — bilingual EN + PT-BR (mandatory)

-Emit a strict JSON object. No prose. No code fence.
+Emit a strict JSON object. No prose. No code fence. Every narrative field
+appears in EN AND in PT-BR (Brazilian Portuguese with UTF-8 accents).

 ```json
 {
  "new_posterior": 0.45,
  "new_confidence_band": "low",
  "delta": 0.05,
-  "rationale": "Concrete prose with [[doc-id/pNNN#cNNNN]] citations.",
-  "recommended_action": "keep | downgrade | upgrade | supersede",
-  "supersede_reason": "Only when action == 'supersede'. Otherwise omit."
+  "rationale":           "EN concrete prose with [[doc-id/pNNN#cNNNN]] citations.",
+  "rationale_pt_br":     "PT-BR prosa concreta com [[doc-id/pNNN#cNNNN]] citações.",
+  "recommended_action":  "keep | downgrade | upgrade | supersede",
+  "supersede_reason":       "EN — only when action == 'supersede'. Otherwise omit.",
+  "supersede_reason_pt_br": "PT-BR — só quando action == 'supersede'. Caso contrário, omita."
 }
 ```

 Constraints:
 - `new_posterior` ∈ [0, 1].
 - `new_confidence_band` MUST match the band thresholds for `new_posterior`.
- `rationale` ≤ 600 chars.
- `supersede_reason` ≤ 280 chars.
+- `rationale` ≤ 1200 chars (per language).
+- `supersede_reason` ≤ 280 chars (per language).
+- A missing `_pt_br` sibling is a hard validation failure.

 If the corpus has NO new evidence since the hypothesis was last reviewed
 (no chunks beyond what was already cited), emit `NO_NEW_EVIDENCE` and
--- a/investigator-runtime/src/detectives/case_writer.ts
+++ b/investigator-runtime/src/detectives/case_writer.ts
@ -24,6 +24,7 @@ const PROMPT_PATH = path.resolve(HERE, "..", "..", "prompts", "case-writer.md");
 export interface CaseWriterTask {
  job_id: string;
  topic: string;
+  topic_pt_br?: string;
  /** When set, restrict to artefacts touching this doc_id (via chunk FK). */
  doc_id?: string;
  lang?: "pt" | "en";
@ -164,15 +165,18 @@ function buildPrompt(
  witnesses: WitnessRow[],
  gaps: GapRow[],
 ): string {
-  const langNote = task.lang === "en"
-    ? "Write the narrative in English."
-    : "Escreva a narrativa em português brasileiro (PT-BR), preservando acentos UTF-8. Cite [[wiki-links]] em inglês como aparecem nos artefatos.";
  return [
-    `# Case folder: ${task.topic}`,
+    `# Case folder`,
+    "",
+    `**Topic (EN).** ${task.topic}`,
+    `**Tópico (PT-BR).** ${task.topic_pt_br ?? task.topic}`,
    "",
    task.doc_id ? `Scoped to document: ${task.doc_id}` : "Scope: all documents",
    "",
-    langNote,
+    "**Bilingual output mandatory.** Write each act in BOTH English and",
+    "Brazilian Portuguese (PT-BR), interleaved per the system-prompt",
+    "structure. UTF-8 accents preserved. Verbatim chunk quotes stay in",
+    "their source language; only the surrounding narration is translated.",
    "",
    "## Artefacts available",
    "",
@ -321,7 +325,7 @@ export async function runCaseWriter(task: CaseWriterTask): Promise<
  if (body_md === null) return { skipped: true, reason: "INSUFFICIENT_ARTEFACTS" };

  return await writeCaseReport({
-    topic, slug, body_md,
+    topic, topic_pt_br: task.topic_pt_br, slug, body_md,
    meta: {
      n_evidence: evidence.length,
      n_hypotheses: hypotheses.length,
--- a/investigator-runtime/src/detectives/dupin.ts
+++ b/investigator-runtime/src/detectives/dupin.ts
@ -66,8 +66,9 @@ function buildPrompt(task: DupinTask, hits: SearchHit[], lang: "pt" | "en"): str
    "Inspect the chunks for pairs (or small groups) that cannot both be true.",
    "Emit at most 3 contradictions. Each must cite ≥ 2 distinct chunk_ids.",
    "Emit the JSON array exactly as specified by the system prompt — no prose,",
-    "no code fence, no preamble. If no genuine contradiction exists,",
-    "emit the literal word `NO_CONTRADICTIONS`.",
+    "no code fence, no preamble. **Bilingual is mandatory:** every narrative",
+    "field (topic, notes, statement) appears in both EN and PT-BR. If no",
+    "genuine contradiction exists, emit the literal word `NO_CONTRADICTIONS`.",
  ].join("\n");
 }

@ -94,9 +95,10 @@ function coercePositions(raw: unknown): ContradictionPosition[] {
    const doc_id = typeof o.doc_id === "string" ? o.doc_id.trim() : "";
    const chunk_id = typeof o.chunk_id === "string" ? o.chunk_id.trim() : "";
    const statement = typeof o.statement === "string" ? o.statement.trim() : "";
-    if (!doc_id || !chunk_id || !statement) continue;
+    const statement_pt_br = typeof o.statement_pt_br === "string" ? o.statement_pt_br.trim() : "";
+    if (!doc_id || !chunk_id || !statement || !statement_pt_br) continue;
    out.push({
-      doc_id, chunk_id, statement,
+      doc_id, chunk_id, statement, statement_pt_br,
      stance: typeof o.stance === "string" ? o.stance.trim() : undefined,
    });
  }
@ -179,12 +181,14 @@ export async function runDupin(task: DupinTask): Promise<
    if (!raw || typeof raw !== "object") continue;
    const o = raw as Record<string, unknown>;
    const topic = typeof o.topic === "string" ? o.topic.trim() : "";
+    const topic_pt_br = typeof o.topic_pt_br === "string" ? o.topic_pt_br.trim() : "";
    const positions = coercePositions(o.positions);
-    if (!topic || positions.length < 2) continue;
+    if (!topic || !topic_pt_br || positions.length < 2) continue;
    const args: WriteContradictionArgs = {
-      topic,
+      topic, topic_pt_br,
      positions,
      notes: typeof o.notes === "string" ? o.notes.trim() : undefined,
+      notes_pt_br: typeof o.notes_pt_br === "string" ? o.notes_pt_br.trim() : undefined,
      resolution_status: o.resolution_status === "resolved"
        ? "resolved"
        : o.resolution_status === "irreconcilable"
--- a/investigator-runtime/src/detectives/holmes.ts
+++ b/investigator-runtime/src/detectives/holmes.ts
@ -28,6 +28,9 @@ const PROMPT_PATH = path.resolve(HERE, "..", "..", "prompts", "holmes.md");
 export interface HolmesTask {
  job_id: string;
  question: string;
+  /** Optional PT-BR mirror of the question. If omitted, the EN one is used
+   *  for both sides until the model emits PT-BR output. */
+  question_pt_br?: string;
  /** Optional scope narrowing — restrict the search to one doc / entity. */
  doc_id?: string;
  lang?: "pt" | "en";
@ -54,10 +57,12 @@ function renderChunkBlock(hits: SearchHit[], lang: "pt" | "en"): string {

 function buildPrompt(task: HolmesTask, hits: SearchHit[], lang: "pt" | "en"): string {
  const block = renderChunkBlock(hits, lang);
+  const ptQ = task.question_pt_br?.trim();
  return [
    `# Question to investigate`,
    "",
-    task.question,
+    `**EN.** ${task.question}`,
+    ptQ ? `**PT-BR.** ${ptQ}` : null,
    "",
    `## Corpus shortlist (${hits.length} chunks${task.doc_id ? `, scoped to ${task.doc_id}` : ""})`,
    "",
@ -66,11 +71,13 @@ function buildPrompt(task: HolmesTask, hits: SearchHit[], lang: "pt" | "en"): st
    "## Your task",
    "",
    "Build 2-3 rival hypotheses about the question above. Each must cite at",
-    "least one chunk via [[doc-id/pNNN#cNNNN]] in argument_for and",
-    "argument_against. Assign priors + posteriors summing roughly to 1.0.",
-    "Emit the JSON array exactly as specified by the system prompt — no prose,",
-    "no code fence, no preamble.",
-  ].join("\n");
+    "least one chunk via [[doc-id/pNNN#cNNNN]] in both argument_for and",
+    "argument_against (EN) and in argument_for_pt_br and",
+    "argument_against_pt_br (PT-BR). Assign priors + posteriors summing",
+    "roughly to 1.0. Emit the JSON array exactly as specified by the system",
+    "prompt — no prose, no code fence, no preamble. **Bilingual is mandatory:",
+    "every narrative field appears in both EN and PT-BR.**",
+  ].filter(Boolean).join("\n");
 }

 function extractJsonArray(text: string): unknown[] | null {
@ -143,20 +150,26 @@ export async function runHolmes(task: HolmesTask): Promise<

  const out: Array<{ hypothesis_id: string; case_file: string }> = [];
  for (const raw of arr.slice(0, 3)) {
+    const r = raw as Record<string, unknown>;
+    const strOrUndef = (k: string): string | undefined =>
+      typeof r[k] === "string" && (r[k] as string).trim().length > 0
+        ? (r[k] as string).trim() : undefined;
    const args: WriteHypothesisArgs = {
-      question: task.question,
-      position: String((raw as { position?: unknown }).position ?? "").trim(),
-      argument_for: typeof (raw as { argument_for?: unknown }).argument_for === "string"
-        ? (raw as { argument_for: string }).argument_for : undefined,
-      argument_against: typeof (raw as { argument_against?: unknown }).argument_against === "string"
-        ? (raw as { argument_against: string }).argument_against : undefined,
-      prior: Number((raw as { prior?: unknown }).prior),
-      posterior: Number((raw as { posterior?: unknown }).posterior),
-      confidence_band: (raw as { confidence_band?: WriteHypothesisArgs["confidence_band"] }).confidence_band,
-      evidence_refs: Array.isArray((raw as { evidence_refs?: unknown }).evidence_refs)
-        ? (raw as { evidence_refs: Array<{ evidence_id?: string; supports?: boolean; weight?: number }> }).evidence_refs
-            .filter((r): r is { evidence_id: string; supports?: boolean; weight?: number } =>
-              typeof r?.evidence_id === "string" && r.evidence_id.length > 0)
+      question:                task.question,
+      question_pt_br:          task.question_pt_br ?? task.question,
+      position:                String(r.position ?? "").trim(),
+      position_pt_br:          strOrUndef("position_pt_br"),
+      argument_for:            strOrUndef("argument_for"),
+      argument_for_pt_br:      strOrUndef("argument_for_pt_br"),
+      argument_against:        strOrUndef("argument_against"),
+      argument_against_pt_br:  strOrUndef("argument_against_pt_br"),
+      prior:                   Number(r.prior),
+      posterior:               Number(r.posterior),
+      confidence_band:         r.confidence_band as WriteHypothesisArgs["confidence_band"],
+      evidence_refs: Array.isArray(r.evidence_refs)
+        ? (r.evidence_refs as Array<{ evidence_id?: string; supports?: boolean; weight?: number }>)
+            .filter((x): x is { evidence_id: string; supports?: boolean; weight?: number } =>
+              typeof x?.evidence_id === "string" && x.evidence_id.length > 0)
        : [],
    };
    if (!args.position) continue;
--- a/investigator-runtime/src/detectives/poirot.ts
+++ b/investigator-runtime/src/detectives/poirot.ts
@ -90,8 +90,10 @@ function buildPrompt(
    "",
    "Produce the structured witness analysis as specified by the system",
    "prompt. Cite chunk_ids from the shortlist above in",
-    "`corroboration_refs`. If the shortlist is too thin to ground an",
-    "honest assessment, emit `INSUFFICIENT_TESTIMONY`.",
+    "`corroboration_refs`. **Bilingual is mandatory:** access_to_event,",
+    "bias_notes, and verdict each appear in both EN and PT-BR. If the",
+    "shortlist is too thin to ground an honest assessment, emit",
+    "`INSUFFICIENT_TESTIMONY`.",
  ].filter(Boolean).join("\n");
 }

@ -224,16 +226,22 @@ export async function runPoirot(task: PoirotTask): Promise<
      ? obj.credibility as "high" | "medium" | "low" | "speculation"
      : "speculation";

+  const str = (k: string): string =>
+    typeof obj[k] === "string" ? (obj[k] as string).trim() : "";
  const args: WriteWitnessAnalysisArgs = {
    person_entity_pk: entity_pk,
    credibility,
-    access_to_event: typeof obj.access_to_event === "string" ? obj.access_to_event.trim() : "",
-    bias_notes: typeof obj.bias_notes === "string" ? obj.bias_notes.trim() : "",
-    corroboration_refs: coerceCorroboration(obj.corroboration_refs),
-    verdict: typeof obj.verdict === "string" ? obj.verdict.trim() : "",
+    access_to_event:       str("access_to_event"),
+    access_to_event_pt_br: str("access_to_event_pt_br"),
+    bias_notes:            str("bias_notes"),
+    bias_notes_pt_br:      str("bias_notes_pt_br"),
+    corroboration_refs:    coerceCorroboration(obj.corroboration_refs),
+    verdict:               str("verdict"),
+    verdict_pt_br:         str("verdict_pt_br"),
  };
-  if (!args.access_to_event || !args.bias_notes || !args.verdict) {
-    return { skipped: true, reason: "incomplete_analysis" };
+  if (!args.access_to_event || !args.bias_notes || !args.verdict
+      || !args.access_to_event_pt_br || !args.bias_notes_pt_br || !args.verdict_pt_br) {
+    return { skipped: true, reason: "incomplete_bilingual_analysis" };
  }

  // Pass the shortlist's most-represented doc_id as a fallback for chunk_id
--- a/investigator-runtime/src/detectives/schneier.ts
+++ b/investigator-runtime/src/detectives/schneier.ts
@ -85,8 +85,10 @@ function buildPrompt(h: HypothesisRow, evidence: EvidenceRow[]): string {
    "",
    "Red-team the hypothesis. Find what the author didn't address. Emit the",
    "JSON object exactly as specified by the system prompt — no prose, no",
-    "code fence, no preamble. If the hypothesis is too thin to attack,",
-    "emit the literal word `INSUFFICIENT_HYPOTHESIS`.",
+    "code fence, no preamble. **Bilingual is mandatory:** every narrative",
+    "field appears in both EN and PT-BR with matching array lengths. If",
+    "the hypothesis is too thin to attack, emit the literal word",
+    "`INSUFFICIENT_HYPOTHESIS`.",
  ].join("\n");
 }

@ -176,16 +178,21 @@ export async function runSchneier(task: SchneierTask): Promise<
  const args: RedTeamReviewArgs = {
    hypothesis_id: h.hypothesis_id,
    severity,
-    hidden_assumptions: coerceStringArray(obj.hidden_assumptions),
-    failure_modes: coerceStringArray(obj.failure_modes),
-    alternative_explanations: coerceStringArray(obj.alternative_explanations),
-    recommended_tests: coerceStringArray(obj.recommended_tests),
+    hidden_assumptions:             coerceStringArray(obj.hidden_assumptions),
+    hidden_assumptions_pt_br:       coerceStringArray(obj.hidden_assumptions_pt_br),
+    failure_modes:                  coerceStringArray(obj.failure_modes),
+    failure_modes_pt_br:            coerceStringArray(obj.failure_modes_pt_br),
+    alternative_explanations:       coerceStringArray(obj.alternative_explanations),
+    alternative_explanations_pt_br: coerceStringArray(obj.alternative_explanations_pt_br),
+    recommended_tests:              coerceStringArray(obj.recommended_tests),
+    recommended_tests_pt_br:        coerceStringArray(obj.recommended_tests_pt_br),
    verdict_one_sentence: typeof obj.verdict_one_sentence === "string"
-      ? obj.verdict_one_sentence.trim()
-      : "",
+      ? obj.verdict_one_sentence.trim() : "",
+    verdict_one_sentence_pt_br: typeof obj.verdict_one_sentence_pt_br === "string"
+      ? obj.verdict_one_sentence_pt_br.trim() : "",
  };
-  if (!args.verdict_one_sentence) {
-    return { skipped: true, reason: "no_verdict" };
+  if (!args.verdict_one_sentence || !args.verdict_one_sentence_pt_br) {
+    return { skipped: true, reason: "no_verdict_bilingual" };
  }

  return await writeRedTeamReview(args, {
--- a/investigator-runtime/src/detectives/taleb.ts
+++ b/investigator-runtime/src/detectives/taleb.ts
@ -61,8 +61,9 @@ function buildPrompt(task: TalebTask, hits: SearchHit[], lang: "pt" | "en"): str
    "",
    "Identify AT MOST 3 outliers per the system prompt rules. State the",
    "dominant_model first, then the chunk that violates it. Emit the",
-    "JSON array exactly as specified — no prose, no code fence. If",
-    "nothing genuinely stands out, emit `NO_OUTLIERS`.",
+    "JSON array exactly as specified — no prose, no code fence.",
+    "**Bilingual is mandatory:** every narrative field appears in both",
+    "EN and PT-BR. If nothing genuinely stands out, emit `NO_OUTLIERS`.",
  ].join("\n");
 }

@ -144,21 +145,36 @@ export async function runTaleb(task: TalebTask): Promise<
  for (const raw of arr.slice(0, 3)) {
    if (!raw || typeof raw !== "object") continue;
    const o = raw as Record<string, unknown>;
+    const str = (k: string): string =>
+      typeof o[k] === "string" ? (o[k] as string).trim() : "";
    const args: WriteOutlierGapArgs = {
-      title: typeof o.title === "string" ? o.title.trim() : "",
-      doc_id: typeof o.doc_id === "string" ? o.doc_id.trim() : "",
-      chunk_id: typeof o.chunk_id === "string" ? o.chunk_id.trim() : "",
-      dominant_model: typeof o.dominant_model === "string" ? o.dominant_model.trim() : "",
-      why_surprising: typeof o.why_surprising === "string" ? o.why_surprising.trim() : "",
-      what_it_implies: typeof o.what_it_implies === "string" ? o.what_it_implies.trim() : "",
-      suggested_next_move: typeof o.suggested_next_move === "string" ? o.suggested_next_move.trim() : "",
+      title:                      str("title"),
+      title_pt_br:                str("title_pt_br"),
+      doc_id:                     str("doc_id"),
+      chunk_id:                   str("chunk_id"),
+      dominant_model:             str("dominant_model"),
+      dominant_model_pt_br:       str("dominant_model_pt_br"),
+      why_surprising:             str("why_surprising"),
+      why_surprising_pt_br:       str("why_surprising_pt_br"),
+      what_it_implies:            str("what_it_implies"),
+      what_it_implies_pt_br:      str("what_it_implies_pt_br"),
+      suggested_next_move:        str("suggested_next_move"),
+      suggested_next_move_pt_br:  str("suggested_next_move_pt_br"),
    };
-    if (!args.title || !args.doc_id || !args.chunk_id || !args.dominant_model
-        || !args.why_surprising || !args.what_it_implies || !args.suggested_next_move) {
+    const argsAny = args as unknown as Record<string, unknown>;
+    const missing = [
+      "title", "title_pt_br", "doc_id", "chunk_id",
+      "dominant_model", "dominant_model_pt_br",
+      "why_surprising", "why_surprising_pt_br",
+      "what_it_implies", "what_it_implies_pt_br",
+      "suggested_next_move", "suggested_next_move_pt_br",
+    ].filter((k) => !argsAny[k]);
+    if (missing.length > 0) {
      await audit({
        event: "write_outlier_gap_failed",
        job_id: task.job_id, detective: "taleb@detective",
-        reason: "incomplete_outlier", title: args.title.slice(0, 120),
+        reason: "incomplete_bilingual_outlier",
+        missing, title: args.title.slice(0, 120),
      });
      continue;
    }
--- a/investigator-runtime/src/detectives/tetlock.ts
+++ b/investigator-runtime/src/detectives/tetlock.ts
@ -127,8 +127,10 @@ function buildPrompt(
    "## Your task",
    "",
    "Recompute the posterior honestly. Emit the JSON object exactly as",
-    "specified by the system prompt. If there is NO new chunk to move the",
-    "posterior on, emit `NO_NEW_EVIDENCE`.",
+    "specified by the system prompt. **Bilingual is mandatory:** rationale",
+    "appears in both EN and PT-BR; supersede_reason (when present) also",
+    "bilingual. If there is NO new chunk to move the posterior on, emit",
+    "`NO_NEW_EVIDENCE`.",
  ].join("\n");
 }

@ -242,15 +244,19 @@ export async function runTetlock(task: TetlockTask): Promise<
    new_posterior,
    new_confidence_band: bandFromPosterior(new_posterior),
    delta,
-    rationale: typeof obj.rationale === "string" ? obj.rationale.trim() : "",
+    rationale:               typeof obj.rationale === "string" ? obj.rationale.trim() : "",
+    rationale_pt_br:         typeof obj.rationale_pt_br === "string" ? obj.rationale_pt_br.trim() : "",
    recommended_action: action,
-    supersede_reason: typeof obj.supersede_reason === "string" ? obj.supersede_reason.trim() : undefined,
+    supersede_reason:        typeof obj.supersede_reason === "string" ? obj.supersede_reason.trim() : undefined,
+    supersede_reason_pt_br:  typeof obj.supersede_reason_pt_br === "string" ? obj.supersede_reason_pt_br.trim() : undefined,
    old_posterior,
    old_confidence_band: h.confidence_band,
  };
-  if (!args.rationale) return { skipped: true, reason: "no_rationale" };
-  if (action === "supersede" && !args.supersede_reason) {
-    return { skipped: true, reason: "supersede_reason_missing" };
+  if (!args.rationale || !args.rationale_pt_br) {
+    return { skipped: true, reason: "no_rationale_bilingual" };
+  }
+  if (action === "supersede" && (!args.supersede_reason || !args.supersede_reason_pt_br)) {
+    return { skipped: true, reason: "supersede_reason_missing_bilingual" };
  }

  return await writeCalibration(args, {
--- a/investigator-runtime/src/orchestrator.ts
+++ b/investigator-runtime/src/orchestrator.ts
@ -55,12 +55,14 @@ export async function dispatch(job: InvestigationJob, workerId: string): Promise
        break;
      }
      case "hypothesis_tournament": {
-        // Payload: { question, doc_id?, lang?, context_chunks? }
+        // Payload: { question, question_pt_br?, doc_id?, lang?, context_chunks? }
        const question = String(job.payload.question ?? "").trim();
        if (!question) throw new Error("hypothesis_tournament requires payload.question");
        const task: HolmesTask = {
          job_id: job.job_id,
          question,
+          question_pt_br: typeof job.payload.question_pt_br === "string"
+            ? job.payload.question_pt_br.trim() : undefined,
          doc_id: typeof job.payload.doc_id === "string" ? job.payload.doc_id : undefined,
          lang: job.payload.lang === "en" ? "en" : "pt",
          context_chunks: typeof job.payload.context_chunks === "number" ? job.payload.context_chunks : undefined,
@ -74,11 +76,13 @@ export async function dispatch(job: InvestigationJob, workerId: string): Promise
        break;
      }
      case "case_report": {
-        // Payload: { topic, doc_id?, slug?, lang? }
+        // Payload: { topic, topic_pt_br?, doc_id?, slug?, lang? }
        const topic = String(job.payload.topic ?? "").trim();
        if (!topic) throw new Error("case_report requires payload.topic");
        const task: CaseWriterTask = {
          job_id: job.job_id, topic,
+          topic_pt_br: typeof job.payload.topic_pt_br === "string"
+            ? job.payload.topic_pt_br.trim() : undefined,
          doc_id: typeof job.payload.doc_id === "string" ? job.payload.doc_id : undefined,
          slug: typeof job.payload.slug === "string" ? job.payload.slug : undefined,
          lang: job.payload.lang === "en" ? "en" : "pt",
--- a/investigator-runtime/src/tools/write_calibration.ts
+++ b/investigator-runtime/src/tools/write_calibration.ts
@ -18,8 +18,10 @@ export interface WriteCalibrationArgs {
  new_confidence_band: "high" | "medium" | "low" | "speculation";
  delta: number;
  rationale: string;
+  rationale_pt_br?: string;
  recommended_action: "keep" | "downgrade" | "upgrade" | "supersede";
  supersede_reason?: string;
+  supersede_reason_pt_br?: string;
  /** previous posterior captured at call time — used in the case-file row. */
  old_posterior: number | null;
  old_confidence_band: string | null;
@ -41,6 +43,7 @@ function bandFromPosterior(p: number): "high" | "medium" | "low" | "speculation"

 function buildSection(args: WriteCalibrationArgs, ctx: WriteCalibrationContext): string {
  const ts = new Date().toISOString();
+  const ptRationale = args.rationale_pt_br ?? args.rationale;
  const rows = [
    `### ${ts} — ${args.recommended_action}`,
    "",
@ -52,10 +55,13 @@ function buildSection(args: WriteCalibrationArgs, ctx: WriteCalibrationContext):
    `| band | ${args.old_confidence_band ?? "—"} | **${args.new_confidence_band}** |`,
    `| delta | — | ${args.delta >= 0 ? "+" : ""}${args.delta.toFixed(3)} |`,
    "",
-    `**Rationale.** ${args.rationale}`,
+    `**Rationale (EN).** ${args.rationale}`,
+    "",
+    `**Justificativa (PT-BR).** ${ptRationale}`,
  ];
  if (args.recommended_action === "supersede" && args.supersede_reason) {
-    rows.push("", `**Supersede reason.** ${args.supersede_reason}`);
+    rows.push("", `**Supersede reason (EN).** ${args.supersede_reason}`);
+    rows.push("", `**Motivo da substituição (PT-BR).** ${args.supersede_reason_pt_br ?? args.supersede_reason}`);
  }
  rows.push("");
  return rows.join("\n");
@ -86,10 +92,10 @@ export async function writeCalibration(
  // Force the band to match the posterior — Tetlock can mis-label.
  body.new_confidence_band = expectedBand;
  if (!body.rationale?.trim()) throw new Error("rationale required");
-  // Soft cap: 1200 chars. Tetlock often writes 600-800 of substantive
-  // reasoning + chunk citations; the prompt asks for ≤ 600 but a 2× slack
-  // beats failing the job on an honest analysis.
+  if (!body.rationale_pt_br?.trim()) throw new Error("rationale_pt_br required (bilingual contract)");
+  // Soft cap: 1200 chars per language.
  if (body.rationale.length > 1200) throw new Error(`rationale too long (${body.rationale.length} > 1200)`);
+  if (body.rationale_pt_br.length > 1200) throw new Error(`rationale_pt_br too long`);

  const action = body.recommended_action;
  if (!["keep", "downgrade", "upgrade", "supersede"].includes(action)) {
--- a/investigator-runtime/src/tools/write_case_report.ts
+++ b/investigator-runtime/src/tools/write_case_report.ts
@ -13,6 +13,7 @@ import { env } from "../lib/env";

 export interface WriteCaseReportArgs {
  topic: string;
+  topic_pt_br?: string;
  slug: string;
  body_md: string;
  meta: {
@ -36,6 +37,7 @@ function renderFrontmatter(args: WriteCaseReportArgs, ctx: WriteCaseReportContex
    `schema_version: "0.1.0"`,
    `type: case_report`,
    `topic: ${JSON.stringify(args.topic)}`,
+    `topic_pt_br: ${JSON.stringify(args.topic_pt_br ?? args.topic)}`,
    `slug: ${args.slug}`,
    `created_by: ${ctx.detective}`,
    `job_id: ${ctx.job_id}`,
--- a/investigator-runtime/src/tools/write_contradiction.ts
+++ b/investigator-runtime/src/tools/write_contradiction.ts
@ -27,14 +27,17 @@ export interface ContradictionPosition {
  chunk_id: string;
  /** The verbatim or paraphrased claim that puts this chunk on this side. */
  statement: string;
+  statement_pt_br?: string;
  /** Optional weight or stance label (e.g. "asserts", "denies"). */
  stance?: string;
 }

 export interface WriteContradictionArgs {
  topic: string;
+  topic_pt_br?: string;
  positions: ContradictionPosition[];
  notes?: string;
+  notes_pt_br?: string;
  resolution_status?: "open" | "resolved" | "irreconcilable";
 }

@ -46,6 +49,7 @@ export interface WriteContradictionContext {
 interface ResolvedPosition extends ContradictionPosition {
  chunk_pk: number;
  page: number;
+  statement_pt_br: string;
 }

 /**
@ -90,7 +94,9 @@ function renderMd(
    return [
      `### Position ${i + 1}${p.stance ? ` — ${p.stance}` : ""}`,
      "",
-      `> ${p.statement}`,
+      `**(EN)** > ${p.statement}`,
+      "",
+      `**(PT-BR)** > ${p.statement_pt_br}`,
      "",
      `Source: [[${p.doc_id}/p${pageStr}#${p.chunk_id}]]`,
    ].join("\n");
@ -101,16 +107,21 @@ function renderMd(
    "",
    `# Contradiction ${id}`,
    "",
-    `**Topic.** ${body.topic}`,
+    `**Topic (EN).** ${body.topic}`,
+    `**Tópico (PT-BR).** ${body.topic_pt_br ?? body.topic}`,
    "",
    "## Positions in tension",
    "",
    positionBlocks.join("\n\n"),
    "",
-    "## Notes",
+    "## Notes (EN)",
    "",
    body.notes || "_(no commentary recorded)_",
    "",
+    "## Notas (PT-BR)",
+    "",
+    body.notes_pt_br || "_(sem comentário registrado)_",
+    "",
  ].join("\n");
 }

@ -119,12 +130,16 @@ export async function writeContradiction(
  ctx: WriteContradictionContext,
 ): Promise<{ contradiction_id: string; case_file: string }> {
  if (!body.topic?.trim()) throw new Error("topic required");
+  if (!body.topic_pt_br?.trim()) throw new Error("topic_pt_br required (bilingual contract)");
  if (!Array.isArray(body.positions) || body.positions.length < 2) {
    throw new Error("at least 2 positions required");
  }
  if (body.notes && body.notes.length > 4000) {
    throw new Error(`notes too long (${body.notes.length} > 4000)`);
  }
+  if (body.notes && !body.notes_pt_br?.trim()) {
+    throw new Error("notes_pt_br required when notes set (bilingual contract)");
+  }

  const resolved: ResolvedPosition[] = [];
  for (const p of body.positions) {
@ -134,6 +149,9 @@ export async function writeContradiction(
    if (!p?.statement?.trim()) {
      throw new Error(`position ${p.doc_id}/${p.chunk_id} missing statement`);
    }
+    if (!p?.statement_pt_br?.trim()) {
+      throw new Error(`position ${p.doc_id}/${p.chunk_id} missing statement_pt_br (bilingual contract)`);
+    }
    const chunk = await resolveChunk(p.doc_id, p.chunk_id);
    if (!chunk) {
      throw new Error(`chunk ${p.doc_id}/${p.chunk_id} not found`);
@ -141,6 +159,7 @@ export async function writeContradiction(
    resolved.push({
      ...p,
      statement: p.statement.trim(),
+      statement_pt_br: p.statement_pt_br.trim(),
      chunk_pk: chunk.chunk_pk,
      page: chunk.page,
    });
@ -160,20 +179,24 @@ export async function writeContradiction(
    chunk_id: p.chunk_id,
    page: p.page,
    statement: p.statement,
+    statement_pt_br: p.statement_pt_br,
    stance: p.stance ?? null,
  }));

  await query(
    `INSERT INTO public.contradictions
-       (contradiction_id, topic, chunks, detected_by, resolution_status, notes)
-     VALUES ($1, $2, $3::jsonb, $4, $5, $6)`,
+       (contradiction_id, topic, topic_pt_br, chunks, detected_by,
+        resolution_status, notes, notes_pt_br)
+     VALUES ($1, $2, $3, $4::jsonb, $5, $6, $7, $8)`,
    [
      contradiction_id,
      body.topic.trim(),
+      body.topic_pt_br!.trim(),
      JSON.stringify(chunkPayload),
      ctx.detective,
      body.resolution_status ?? "open",
      body.notes ?? null,
+      body.notes_pt_br ?? null,
    ],
  );

--- a/investigator-runtime/src/tools/write_hypothesis.ts
+++ b/investigator-runtime/src/tools/write_hypothesis.ts
@ -25,9 +25,13 @@ export interface EvidenceRef {

 export interface WriteHypothesisArgs {
  question: string;
+  question_pt_br?: string;
  position: string;
+  position_pt_br?: string;
  argument_for?: string;
+  argument_for_pt_br?: string;
  argument_against?: string;
+  argument_against_pt_br?: string;
  prior?: number;
  posterior?: number;
  confidence_band?: "high" | "medium" | "low" | "speculation";
@ -70,23 +74,39 @@ function renderMd(id: string, body: WriteHypothesisArgs, ctx: WriteHypothesisCon
    `created_at: ${new Date().toISOString()}`,
    "---",
  ].filter(Boolean).join("\n");
+
+  const ptQuestion = body.question_pt_br ?? body.question;
+  const ptPosition = body.position_pt_br ?? body.position;
+  const ptFor      = body.argument_for_pt_br ?? body.argument_for;
+  const ptAgainst  = body.argument_against_pt_br ?? body.argument_against;
+
  return [
    fm,
    "",
    `# Hypothesis ${id}`,
    "",
-    `**Question.** ${body.question}`,
+    `**Question (EN).** ${body.question}`,
+    `**Pergunta (PT-BR).** ${ptQuestion}`,
    "",
-    `**Position.** ${body.position}`,
+    `**Position (EN).** ${body.position}`,
+    `**Posição (PT-BR).** ${ptPosition}`,
    "",
-    "## Argument for",
+    "## Argument for (EN)",
    "",
    body.argument_for || "_(none recorded — speculation)_",
    "",
-    "## Argument against",
+    "## Argumento a favor (PT-BR)",
+    "",
+    ptFor || "_(nenhum registrado — especulação)_",
+    "",
+    "## Argument against (EN)",
    "",
    body.argument_against || "_(none recorded — no counter-argument framed yet)_",
    "",
+    "## Argumento contra (PT-BR)",
+    "",
+    ptAgainst || "_(nenhum registrado — sem contra-argumento ainda)_",
+    "",
    "## Evidence",
    "",
    evRefs || "_(none linked yet — Locard chain pending)_",
@ -100,6 +120,13 @@ export async function writeHypothesis(
 ): Promise<{ hypothesis_id: string; case_file: string }> {
  if (!body.question?.trim()) throw new Error("question required");
  if (!body.position?.trim()) throw new Error("position required");
+  if (!body.position_pt_br?.trim()) throw new Error("position_pt_br required (bilingual contract)");
+  if (body.argument_for && !body.argument_for_pt_br?.trim()) {
+    throw new Error("argument_for_pt_br required when argument_for is set (bilingual contract)");
+  }
+  if (body.argument_against && !body.argument_against_pt_br?.trim()) {
+    throw new Error("argument_against_pt_br required when argument_against is set (bilingual contract)");
+  }

  const prior = clamp01(body.prior);
  const posterior = clamp01(body.posterior);
@ -128,12 +155,16 @@ export async function writeHypothesis(
  const hypothesis_id = await allocate.hypothesisId();
  await query(
    `INSERT INTO public.hypotheses
-       (hypothesis_id, question, position, argument_for, argument_against,
+       (hypothesis_id, question, question_pt_br, position, position_pt_br,
+        argument_for, argument_for_pt_br, argument_against, argument_against_pt_br,
        evidence_refs, prior, posterior, confidence_band, status, created_by)
-     VALUES ($1,$2,$3,$4,$5,$6::jsonb,$7,$8,$9,$10,$11)`,
+     VALUES ($1,$2,$3,$4,$5,$6,$7,$8,$9,$10::jsonb,$11,$12,$13,$14,$15)`,
    [
-      hypothesis_id, body.question, body.position,
-      body.argument_for ?? null, body.argument_against ?? null,
+      hypothesis_id,
+      body.question, body.question_pt_br ?? body.question,
+      body.position, body.position_pt_br,
+      body.argument_for ?? null, body.argument_for_pt_br ?? null,
+      body.argument_against ?? null, body.argument_against_pt_br ?? null,
      JSON.stringify(refs),
      prior, posterior, band, body.status ?? "open",
      ctx.detective,
--- a/investigator-runtime/src/tools/write_outlier_gap.ts
+++ b/investigator-runtime/src/tools/write_outlier_gap.ts
@ -16,12 +16,17 @@ import { query, queryOne } from "../lib/pg";

 export interface WriteOutlierGapArgs {
  title: string;
+  title_pt_br?: string;
  doc_id: string;
  chunk_id: string;
  dominant_model: string;
+  dominant_model_pt_br?: string;
  why_surprising: string;
+  why_surprising_pt_br?: string;
  what_it_implies: string;
+  what_it_implies_pt_br?: string;
  suggested_next_move: string;
+  suggested_next_move_pt_br?: string;
 }

 export interface WriteOutlierGapContext {
@ -61,25 +66,42 @@ function renderMd(
    fm,
    "",
    `# Outlier ${id} — ${body.title}`,
+    `**Título (PT-BR).** ${body.title_pt_br ?? body.title}`,
    "",
    `**Source.** [[${body.doc_id}/p${pageStr}#${body.chunk_id}]]`,
    "",
-    "## Dominant model",
+    "## Dominant model (EN)",
    "",
    body.dominant_model,
    "",
-    "## Why surprising",
+    "## Modelo dominante (PT-BR)",
+    "",
+    body.dominant_model_pt_br ?? body.dominant_model,
+    "",
+    "## Why surprising (EN)",
    "",
    body.why_surprising,
    "",
-    "## What it implies",
+    "## Por que é surpreendente (PT-BR)",
+    "",
+    body.why_surprising_pt_br ?? body.why_surprising,
+    "",
+    "## What it implies (EN)",
    "",
    body.what_it_implies,
    "",
-    "## Suggested next move",
+    "## O que implica (PT-BR)",
+    "",
+    body.what_it_implies_pt_br ?? body.what_it_implies,
+    "",
+    "## Suggested next move (EN)",
    "",
    body.suggested_next_move,
    "",
+    "## Próximo passo sugerido (PT-BR)",
+    "",
+    body.suggested_next_move_pt_br ?? body.suggested_next_move,
+    "",
  ].join("\n");
 }

@ -88,12 +110,18 @@ export async function writeOutlierGap(
  ctx: WriteOutlierGapContext,
 ): Promise<{ gap_id: string; case_file: string }> {
  if (!body.title?.trim()) throw new Error("title required");
+  if (!body.title_pt_br?.trim()) throw new Error("title_pt_br required (bilingual contract)");
  if (!body.doc_id?.trim() || !body.chunk_id?.trim()) throw new Error("doc_id + chunk_id required");
  if (!body.dominant_model?.trim()) throw new Error("dominant_model required");
+  if (!body.dominant_model_pt_br?.trim()) throw new Error("dominant_model_pt_br required (bilingual contract)");
  if (!body.why_surprising?.trim()) throw new Error("why_surprising required");
+  if (!body.why_surprising_pt_br?.trim()) throw new Error("why_surprising_pt_br required (bilingual contract)");
  if (!body.what_it_implies?.trim()) throw new Error("what_it_implies required");
+  if (!body.what_it_implies_pt_br?.trim()) throw new Error("what_it_implies_pt_br required (bilingual contract)");
  if (!body.suggested_next_move?.trim()) throw new Error("suggested_next_move required");
+  if (!body.suggested_next_move_pt_br?.trim()) throw new Error("suggested_next_move_pt_br required (bilingual contract)");
  if (body.why_surprising.length > 600) throw new Error(`why_surprising too long`);
+  if (body.why_surprising_pt_br.length > 600) throw new Error(`why_surprising_pt_br too long`);

  const cid = normalizeChunkId(body.chunk_id);
  const chunk = await queryOne<{ chunk_pk: number; page: number }>(
@ -109,21 +137,28 @@ export async function writeOutlierGap(
    doc_id: body.doc_id,
    chunk_id: cid,
    page: chunk.page,
-    dominant_model: body.dominant_model,
-    why_surprising: body.why_surprising,
-    what_it_implies: body.what_it_implies,
-    title: body.title,
+    title:                  body.title,
+    title_pt_br:            body.title_pt_br,
+    dominant_model:         body.dominant_model,
+    dominant_model_pt_br:   body.dominant_model_pt_br,
+    why_surprising:         body.why_surprising,
+    why_surprising_pt_br:   body.why_surprising_pt_br,
+    what_it_implies:        body.what_it_implies,
+    what_it_implies_pt_br:  body.what_it_implies_pt_br,
  };

  await query(
    `INSERT INTO public.gaps
-       (gap_id, description, scope, suggested_next_move, status, created_by)
-     VALUES ($1, $2, $3::jsonb, $4, 'open', $5)`,
+       (gap_id, description, description_pt_br, scope,
+        suggested_next_move, suggested_next_move_pt_br, status, created_by)
+     VALUES ($1, $2, $3, $4::jsonb, $5, $6, 'open', $7)`,
    [
      gap_id,
      body.title,
+      body.title_pt_br,
      JSON.stringify(scope),
      body.suggested_next_move,
+      body.suggested_next_move_pt_br,
      ctx.detective,
    ],
  );
--- a/investigator-runtime/src/tools/write_red_team_review.ts
+++ b/investigator-runtime/src/tools/write_red_team_review.ts
@ -19,10 +19,15 @@ export interface RedTeamReviewArgs {
  hypothesis_id: string;
  severity: "low" | "medium" | "high";
  hidden_assumptions: string[];
+  hidden_assumptions_pt_br?: string[];
  failure_modes: string[];
+  failure_modes_pt_br?: string[];
  alternative_explanations: string[];
+  alternative_explanations_pt_br?: string[];
  recommended_tests: string[];
+  recommended_tests_pt_br?: string[];
  verdict_one_sentence: string;
+  verdict_one_sentence_pt_br?: string;
 }

 export interface RedTeamReviewContext {
@ -34,8 +39,14 @@ const SECTION_MARKER = "## Red-team review";

 function buildSection(args: RedTeamReviewArgs, ctx: RedTeamReviewContext): string {
  const ts = new Date().toISOString();
-  const bullets = (items: string[]): string =>
-    items.length === 0 ? "_(none flagged)_" : items.map((x) => `- ${x}`).join("\n");
+  const bullets = (items: string[], emptyMsg: string): string =>
+    items.length === 0 ? emptyMsg : items.map((x) => `- ${x}`).join("\n");
+  const pt = (a: string[] | undefined, fb: string[]) => a && a.length === fb.length ? a : fb;
+  const ptHidden  = pt(args.hidden_assumptions_pt_br, args.hidden_assumptions);
+  const ptFail    = pt(args.failure_modes_pt_br, args.failure_modes);
+  const ptAlt     = pt(args.alternative_explanations_pt_br, args.alternative_explanations);
+  const ptTests   = pt(args.recommended_tests_pt_br, args.recommended_tests);
+  const ptVerdict = args.verdict_one_sentence_pt_br ?? args.verdict_one_sentence;

  return [
    "",
@ -43,19 +54,32 @@ function buildSection(args: RedTeamReviewArgs, ctx: RedTeamReviewContext): strin
    "",
    `_Reviewed by ${ctx.detective} on ${ts} — job \`${ctx.job_id}\`._`,
    "",
-    `**Verdict.** ${args.verdict_one_sentence}`,
+    `**Verdict (EN).** ${args.verdict_one_sentence}`,
+    `**Veredito (PT-BR).** ${ptVerdict}`,
    "",
-    "### Hidden assumptions",
-    bullets(args.hidden_assumptions),
+    "### Hidden assumptions (EN)",
+    bullets(args.hidden_assumptions, "_(none flagged)_"),
    "",
-    "### Failure modes",
-    bullets(args.failure_modes),
+    "### Premissas ocultas (PT-BR)",
+    bullets(ptHidden, "_(nenhuma sinalizada)_"),
    "",
-    "### Alternative explanations not addressed",
-    bullets(args.alternative_explanations),
+    "### Failure modes (EN)",
+    bullets(args.failure_modes, "_(none flagged)_"),
    "",
-    "### Recommended discriminating tests",
-    bullets(args.recommended_tests),
+    "### Modos de falha (PT-BR)",
+    bullets(ptFail, "_(nenhum sinalizado)_"),
+    "",
+    "### Alternative explanations not addressed (EN)",
+    bullets(args.alternative_explanations, "_(none flagged)_"),
+    "",
+    "### Explicações alternativas não abordadas (PT-BR)",
+    bullets(ptAlt, "_(nenhuma sinalizada)_"),
+    "",
+    "### Recommended discriminating tests (EN)",
+    bullets(args.recommended_tests, "_(none flagged)_"),
+    "",
+    "### Testes discriminantes recomendados (PT-BR)",
+    bullets(ptTests, "_(nenhum sinalizado)_"),
    "",
  ].join("\n");
 }
@ -82,9 +106,15 @@ export async function writeRedTeamReview(
    throw new Error(`bad severity: ${body.severity}`);
  }
  if (!body.verdict_one_sentence?.trim()) throw new Error("verdict_one_sentence required");
+  if (!body.verdict_one_sentence_pt_br?.trim()) {
+    throw new Error("verdict_one_sentence_pt_br required (bilingual contract)");
+  }
  if (body.verdict_one_sentence.length > 280) {
    throw new Error(`verdict too long (${body.verdict_one_sentence.length} > 280)`);
  }
+  if (body.verdict_one_sentence_pt_br.length > 280) {
+    throw new Error(`verdict_pt_br too long (${body.verdict_one_sentence_pt_br.length} > 280)`);
+  }

  // Defensive: cap each array to 5 entries × 240 chars (prompt says ≤ 200 but
  // we leave some slack rather than truncate silently).
--- a/investigator-runtime/src/tools/write_witness_analysis.ts
+++ b/investigator-runtime/src/tools/write_witness_analysis.ts
@ -31,9 +31,12 @@ export interface WriteWitnessAnalysisArgs {
  person_entity_pk: number;
  credibility: "high" | "medium" | "low" | "speculation";
  access_to_event: string;
+  access_to_event_pt_br?: string;
  bias_notes: string;
+  bias_notes_pt_br?: string;
  corroboration_refs: CorroborationRef[];
  verdict: string;
+  verdict_pt_br?: string;
 }

 export interface WriteWitnessAnalysisContext {
@ -106,16 +109,25 @@ function renderMd(
    "",
    `**Credibility.** ${body.credibility}`,
    "",
-    `**Verdict.** ${body.verdict}`,
+    `**Verdict (EN).** ${body.verdict}`,
+    `**Veredito (PT-BR).** ${body.verdict_pt_br ?? body.verdict}`,
    "",
-    "## Access to event",
+    "## Access to event (EN)",
    "",
    body.access_to_event,
    "",
-    "## Bias notes",
+    "## Acesso ao evento (PT-BR)",
+    "",
+    body.access_to_event_pt_br ?? body.access_to_event,
+    "",
+    "## Bias notes (EN)",
    "",
    body.bias_notes,
    "",
+    "## Notas de viés (PT-BR)",
+    "",
+    body.bias_notes_pt_br ?? body.bias_notes,
+    "",
    "## Corroboration chain",
    "",
    refBlocks,
@ -132,11 +144,17 @@ export async function writeWitnessAnalysis(
  const validBand = ["high", "medium", "low", "speculation"].includes(body.credibility);
  if (!validBand) throw new Error(`bad credibility: ${body.credibility}`);
  if (!body.access_to_event?.trim()) throw new Error("access_to_event required");
+  if (!body.access_to_event_pt_br?.trim()) throw new Error("access_to_event_pt_br required (bilingual contract)");
  if (!body.bias_notes?.trim()) throw new Error("bias_notes required");
+  if (!body.bias_notes_pt_br?.trim()) throw new Error("bias_notes_pt_br required (bilingual contract)");
  if (!body.verdict?.trim()) throw new Error("verdict required");
+  if (!body.verdict_pt_br?.trim()) throw new Error("verdict_pt_br required (bilingual contract)");
  if (body.verdict.length > 280) throw new Error(`verdict too long (${body.verdict.length} > 280)`);
+  if (body.verdict_pt_br.length > 280) throw new Error(`verdict_pt_br too long (${body.verdict_pt_br.length} > 280)`);
  if (body.access_to_event.length > 800) throw new Error(`access_to_event too long (${body.access_to_event.length} > 800)`);
+  if (body.access_to_event_pt_br.length > 800) throw new Error(`access_to_event_pt_br too long`);
  if (body.bias_notes.length > 800) throw new Error(`bias_notes too long (${body.bias_notes.length} > 800)`);
+  if (body.bias_notes_pt_br.length > 800) throw new Error(`bias_notes_pt_br too long`);

  // Verify entity exists and is a person.
  const ent = await queryOne<{ canonical_name: string; entity_class: string }>(
@ -159,17 +177,20 @@ export async function writeWitnessAnalysis(
  const witness_id = await allocate.witnessId();
  await query(
    `INSERT INTO public.witnesses
-       (witness_id, person_entity_pk, credibility, access_to_event,
-        bias_notes, corroboration_refs, verdict, created_by)
-     VALUES ($1, $2, $3, $4, $5, $6::jsonb, $7, $8)`,
+       (witness_id, person_entity_pk, credibility,
+        access_to_event, access_to_event_pt_br,
+        bias_notes, bias_notes_pt_br,
+        corroboration_refs, verdict, verdict_pt_br, created_by)
+     VALUES ($1, $2, $3, $4, $5, $6, $7, $8::jsonb, $9, $10, $11)`,
    [
      witness_id, body.person_entity_pk, body.credibility,
-      body.access_to_event, body.bias_notes,
+      body.access_to_event, body.access_to_event_pt_br,
+      body.bias_notes, body.bias_notes_pt_br,
      JSON.stringify(refs.map((r) => ({
        chunk_pk: r.chunk_pk, doc_id: r.doc_id, chunk_id: r.chunk_id,
        page: r.page, supports: r.supports,
      }))),
-      body.verdict, ctx.detective,
+      body.verdict, body.verdict_pt_br, ctx.detective,
    ],
  );

--- a/web/app/api/jobs/[id]/route.ts
+++ b/web/app/api/jobs/[id]/route.ts
@ -46,9 +46,13 @@ interface EvidenceRow {
 interface HypothesisRow {
  hypothesis_id: string;
  question: string | null;
+  question_pt_br: string | null;
  position: string | null;
+  position_pt_br: string | null;
  argument_for: string | null;
+  argument_for_pt_br: string | null;
  argument_against: string | null;
+  argument_against_pt_br: string | null;
  prior: number | null;
  posterior: number | null;
  confidence_band: string | null;
@ -59,9 +63,11 @@ interface HypothesisRow {
 interface ContradictionRow {
  contradiction_id: string;
  topic: string;
+  topic_pt_br: string | null;
  chunks: unknown;
  resolution_status: string | null;
  notes: string | null;
+  notes_pt_br: string | null;
  detected_by: string | null;
 }

@ -71,16 +77,21 @@ interface WitnessRow {
  entity_id: string | null;
  credibility: string | null;
  access_to_event: string | null;
+  access_to_event_pt_br: string | null;
  bias_notes: string | null;
+  bias_notes_pt_br: string | null;
  corroboration_refs: unknown;
  verdict: string | null;
+  verdict_pt_br: string | null;
 }

 interface GapRow {
  gap_id: string;
  description: string;
+  description_pt_br: string | null;
  scope: unknown;
  suggested_next_move: string | null;
+  suggested_next_move_pt_br: string | null;
  status: string;
  created_by: string;
 }
@ -144,7 +155,8 @@ export async function GET(
        : Promise.resolve([] as EvidenceRow[]),
      hypothesisIds.length > 0
        ? pgQuery<HypothesisRow>(
-            `SELECT hypothesis_id, question, position, argument_for, argument_against,
+            `SELECT hypothesis_id, question, question_pt_br, position, position_pt_br,
+                    argument_for, argument_for_pt_br, argument_against, argument_against_pt_br,
                    prior, posterior, confidence_band, status, evidence_refs
               FROM public.hypotheses
              WHERE hypothesis_id = ANY($1::text[])
@ -154,7 +166,8 @@ export async function GET(
        : Promise.resolve([] as HypothesisRow[]),
      contradictionIds.length > 0
        ? pgQuery<ContradictionRow>(
-            `SELECT contradiction_id, topic, chunks, resolution_status, notes, detected_by
+            `SELECT contradiction_id, topic, topic_pt_br, chunks, resolution_status,
+                    notes, notes_pt_br, detected_by
               FROM public.contradictions
              WHERE contradiction_id = ANY($1::text[])
              ORDER BY contradiction_id`,
@ -163,9 +176,10 @@ export async function GET(
        : Promise.resolve([] as ContradictionRow[]),
      witnessIds.length > 0
        ? pgQuery<WitnessRow>(
-            `SELECT w.witness_id, e.canonical_name, e.entity_id,
-                    w.credibility, w.access_to_event, w.bias_notes,
-                    w.corroboration_refs, w.verdict
+            `SELECT w.witness_id, e.canonical_name, e.entity_id, w.credibility,
+                    w.access_to_event, w.access_to_event_pt_br,
+                    w.bias_notes, w.bias_notes_pt_br,
+                    w.corroboration_refs, w.verdict, w.verdict_pt_br
               FROM public.witnesses w
          LEFT JOIN public.entities e ON e.entity_pk = w.person_entity_pk
              WHERE w.witness_id = ANY($1::text[])
@ -175,7 +189,8 @@ export async function GET(
        : Promise.resolve([] as WitnessRow[]),
      gapIds.length > 0
        ? pgQuery<GapRow>(
-            `SELECT gap_id, description, scope, suggested_next_move, status, created_by
+            `SELECT gap_id, description, description_pt_br, scope,
+                    suggested_next_move, suggested_next_move_pt_br, status, created_by
               FROM public.gaps WHERE gap_id = ANY($1::text[]) ORDER BY gap_id`,
            [gapIds],
          )
--- a/web/app/bureau/page.tsx
+++ b/web/app/bureau/page.tsx
@ -22,7 +22,9 @@ interface EvidenceRow {
 interface HypothesisRow {
  hypothesis_id: string;
  question: string;
+  question_pt_br: string | null;
  position: string;
+  position_pt_br: string | null;
  prior: number | string | null;
  posterior: number | string | null;
  confidence_band: string | null;
@ -33,15 +35,18 @@ interface HypothesisRow {
 interface ContradictionRow {
  contradiction_id: string;
  topic: string;
+  topic_pt_br: string | null;
  resolution_status: string;
  chunks: unknown;
 }
 interface GapRow {
  gap_id: string;
  description: string;
+  description_pt_br: string | null;
  scope: unknown;
  status: string;
  suggested_next_move: string | null;
+  suggested_next_move_pt_br: string | null;
 }
 interface WitnessRow {
  witness_id: string;
@ -49,6 +54,7 @@ interface WitnessRow {
  entity_id: string | null;
  credibility: string | null;
  verdict: string | null;
+  verdict_pt_br: string | null;
 }
 interface JobRow {
  job_id: string;
@ -77,8 +83,8 @@ export default async function BureauPage() {
  // All artefacts. Server component — single round per query, no n+1.
  const [hyp, ev, ctr, gap, wit, jobs] = await Promise.all([
    pgQuery<HypothesisRow>(
-      `SELECT hypothesis_id, question, position, prior, posterior, confidence_band,
-              status, reviewed_by, created_at
+      `SELECT hypothesis_id, question, question_pt_br, position, position_pt_br,
+              prior, posterior, confidence_band, status, reviewed_by, created_at
         FROM public.hypotheses ORDER BY created_at DESC LIMIT 100`,
    ).catch(() => []),
    pgQuery<EvidenceRow>(
@ -86,15 +92,17 @@ export default async function BureauPage() {
         FROM public.evidence ORDER BY created_at DESC LIMIT 100`,
    ).catch(() => []),
    pgQuery<ContradictionRow>(
-      `SELECT contradiction_id, topic, resolution_status, chunks
+      `SELECT contradiction_id, topic, topic_pt_br, resolution_status, chunks
         FROM public.contradictions ORDER BY created_at DESC LIMIT 100`,
    ).catch(() => []),
    pgQuery<GapRow>(
-      `SELECT gap_id, description, scope, status, suggested_next_move
+      `SELECT gap_id, description, description_pt_br, scope, status,
+              suggested_next_move, suggested_next_move_pt_br
         FROM public.gaps ORDER BY created_at DESC LIMIT 100`,
    ).catch(() => []),
    pgQuery<WitnessRow>(
-      `SELECT w.witness_id, e.canonical_name, e.entity_id, w.credibility, w.verdict
+      `SELECT w.witness_id, e.canonical_name, e.entity_id, w.credibility,
+              w.verdict, w.verdict_pt_br
         FROM public.witnesses w
    LEFT JOIN public.entities e ON e.entity_pk = w.person_entity_pk
        ORDER BY w.created_at DESC LIMIT 100`,
@ -183,7 +191,7 @@ export default async function BureauPage() {
                    )}
                  </div>
                </div>
-                <div className="text-[13px] text-[#e7ecf3] leading-snug">{h.position}</div>
+                <div className="text-[13px] text-[#e7ecf3] leading-snug">{h.position_pt_br || h.position}</div>
                <div className="text-[10px] font-mono text-[#5a6678] mt-1">
                  prior {prior?.toFixed(2) ?? "—"} → posterior {post?.toFixed(2) ?? "—"}
                  {delta !== null && <span className={delta > 0 ? " text-[#06d6a0]" : delta < 0 ? " text-[#ff6ec7]" : ""}> · Δ {delta >= 0 ? "+" : ""}{delta.toFixed(3)}</span>}
@ -221,7 +229,7 @@ export default async function BureauPage() {
                  <span className="text-[10px] font-mono text-[#5a6678]">{c.contradiction_id}</span>
                  <span className="text-[10px] font-mono text-[#9aa6b8]">{n} positions · {c.resolution_status}</span>
                </div>
-                <div className="text-[13px] text-[#e7ecf3] leading-snug">{c.topic}</div>
+                <div className="text-[13px] text-[#e7ecf3] leading-snug">{c.topic_pt_br || c.topic}</div>
              </div>
            );
          })}
@ -231,8 +239,10 @@ export default async function BureauPage() {
        <Section id="outliers" title="Outliers" color="text-[#ffd23f]">
          {gap.length === 0 ? <Empty /> : gap.map((g) => {
            const s = (g.scope ?? {}) as Record<string, unknown>;
-            const title = (s.title as string) || g.description;
+            const title = (s.title_pt_br as string) || (s.title as string) || g.description_pt_br || g.description;
+            const why = (s.why_surprising_pt_br as string) || (s.why_surprising as string) || null;
            const isOutlier = s.kind === "outlier";
+            const nextMove = g.suggested_next_move_pt_br || g.suggested_next_move;
            return (
              <div key={g.gap_id} className="rounded border border-[rgba(255,210,63,0.25)] bg-[#0d1220] p-3 mb-2">
                <div className="flex items-baseline justify-between gap-2 mb-1">
@ -240,11 +250,11 @@ export default async function BureauPage() {
                  <span className="text-[10px] font-mono text-[#9aa6b8]">{g.status}</span>
                </div>
                <div className="text-[13px] text-[#e7ecf3] leading-snug">{title}</div>
-                {s.why_surprising !== undefined && (
-                  <div className="text-[11px] text-[#cbd2dd] mt-1 leading-relaxed">{String(s.why_surprising)}</div>
+                {why && (
+                  <div className="text-[11px] text-[#cbd2dd] mt-1 leading-relaxed">{why}</div>
                )}
-                {g.suggested_next_move && (
-                  <div className="text-[10px] font-mono text-[#06d6a0] mt-1">→ {g.suggested_next_move}</div>
+                {nextMove && (
+                  <div className="text-[10px] font-mono text-[#06d6a0] mt-1">→ {nextMove}</div>
                )}
              </div>
            );
@ -262,7 +272,7 @@ export default async function BureauPage() {
              <div className="text-[13px] text-[#e7ecf3] font-medium">
                {w.entity_id ? <Link href={`/e/people/${w.entity_id}`} className="hover:underline">{w.canonical_name ?? w.entity_id}</Link> : (w.canonical_name ?? "—")}
              </div>
-              {w.verdict && <blockquote className="text-[12px] text-[#cbd2dd] italic mt-1 border-l-2 border-[#9b5de5] pl-2">{w.verdict}</blockquote>}
+              {(w.verdict_pt_br || w.verdict) && <blockquote className="text-[12px] text-[#cbd2dd] italic mt-1 border-l-2 border-[#9b5de5] pl-2">{w.verdict_pt_br || w.verdict}</blockquote>}
            </div>
          ))}
        </Section>
--- a/web/app/h/[hypothesisId]/page.tsx
+++ b/web/app/h/[hypothesisId]/page.tsx
@ -25,9 +25,13 @@ export const dynamic = "force-dynamic";
 interface HypothesisRow {
  hypothesis_id: string;
  question: string;
+  question_pt_br: string | null;
  position: string;
+  position_pt_br: string | null;
  argument_for: string | null;
+  argument_for_pt_br: string | null;
  argument_against: string | null;
+  argument_against_pt_br: string | null;
  prior: number | string | null;
  posterior: number | string | null;
  confidence_band: string | null;
@ -146,7 +150,8 @@ export default async function HypothesisPage({
  if (!/^H-\d{4}$/.test(hypothesisId)) notFound();

  const rows = await pgQuery<HypothesisRow>(
-    `SELECT hypothesis_id, question, position, argument_for, argument_against,
+    `SELECT hypothesis_id, question, question_pt_br, position, position_pt_br,
+            argument_for, argument_for_pt_br, argument_against, argument_against_pt_br,
            prior, posterior, confidence_band, status, evidence_refs,
            created_by, reviewed_by, created_at, updated_at
       FROM public.hypotheses WHERE hypothesis_id = $1`,
@ -209,8 +214,11 @@ export default async function HypothesisPage({
                {hypothesisId} · created by {h.created_by}
                {h.reviewed_by && <> · reviewed by <span className="text-[#ff3344]">{h.reviewed_by}</span></>}
              </div>
-              <h1 className="text-xl font-mono text-[#e7ecf3] leading-snug">{h.position}</h1>
-              <p className="text-[12px] text-[#9aa6b8] mt-1">Question: {h.question}</p>
+              <h1 className="text-xl font-mono text-[#e7ecf3] leading-snug">{h.position_pt_br || h.position}</h1>
+              {h.position_pt_br && h.position_pt_br !== h.position && (
+                <p className="text-[12px] text-[#5a6678] italic mt-0.5">{h.position}</p>
+              )}
+              <p className="text-[12px] text-[#9aa6b8] mt-1">Pergunta: {h.question_pt_br || h.question}</p>
            </div>
            {h.confidence_band && (
              <span className={`px-2 py-0.5 rounded text-[10px] font-mono uppercase border ${bandTone}`}>
@ -238,8 +246,8 @@ export default async function HypothesisPage({

        {/* Argument grid */}
        <div className="mt-4 grid md:grid-cols-2 gap-3">
-          <ArgumentPanel kind="for" body={h.argument_for} />
-          <ArgumentPanel kind="against" body={h.argument_against} />
+          <ArgumentPanel kind="for" body={h.argument_for_pt_br || h.argument_for} bodyEn={h.argument_for_pt_br ? h.argument_for : null} />
+          <ArgumentPanel kind="against" body={h.argument_against_pt_br || h.argument_against} bodyEn={h.argument_against_pt_br ? h.argument_against : null} />
        </div>

        {/* Evidence chain */}
@ -273,13 +281,20 @@ export default async function HypothesisPage({
  );
 }

-function ArgumentPanel({ kind, body }: { kind: "for" | "against"; body: string | null }) {
+function ArgumentPanel({ kind, body, bodyEn }: { kind: "for" | "against"; body: string | null; bodyEn?: string | null }) {
  const tone = kind === "for" ? "border-[#06d6a0] text-[#06d6a0]" : "border-[#ff6ec7] text-[#ff6ec7]";
-  const label = kind === "for" ? "Argumento a favor" : "Argumento contra";
+  const label = kind === "for" ? "Argumento a favor (PT-BR)" : "Argumento contra (PT-BR)";
+  const enLabel = kind === "for" ? "Argument for (EN)" : "Argument against (EN)";
  return (
    <div className={`rounded-lg border ${tone} bg-[#0d1220] p-4`}>
      <div className={`text-[10px] font-mono uppercase mb-2 ${tone}`}>{label}</div>
      <ArgumentBody text={body ?? ""} />
+      {bodyEn && (
+        <details className="mt-3 pt-3 border-t border-[rgba(127,219,255,0.08)]">
+          <summary className="text-[10px] font-mono text-[#5a6678] cursor-pointer hover:text-[#9aa6b8]">{enLabel}</summary>
+          <div className="mt-2"><ArgumentBody text={bodyEn} /></div>
+        </details>
+      )}
    </div>
  );
 }
--- a/web/components/bureau-snapshot.tsx
+++ b/web/components/bureau-snapshot.tsx
@ -12,10 +12,10 @@ import { QuickLaunch } from "./quick-launch";

 interface CountRow { c: string }
 interface RecentEvidence { evidence_id: string; grade: string; verbatim_excerpt: string; source_page_id: string; confidence_band: string | null }
-interface RecentHypothesis { hypothesis_id: string; position: string; posterior: number | string | null; confidence_band: string | null; reviewed_by: string | null }
-interface RecentContradiction { contradiction_id: string; topic: string; resolution_status: string }
-interface RecentGap { gap_id: string; description: string; scope: unknown }
-interface RecentWitness { witness_id: string; canonical_name: string | null; credibility: string | null; verdict: string | null }
+interface RecentHypothesis { hypothesis_id: string; position: string; position_pt_br: string | null; posterior: number | string | null; confidence_band: string | null; reviewed_by: string | null }
+interface RecentContradiction { contradiction_id: string; topic: string; topic_pt_br: string | null; resolution_status: string }
+interface RecentGap { gap_id: string; description: string; description_pt_br: string | null; scope: unknown }
+interface RecentWitness { witness_id: string; canonical_name: string | null; credibility: string | null; verdict: string | null; verdict_pt_br: string | null }
 interface RecentJob { job_id: string; kind: string; status: string; created_at: string; payload: Record<string, unknown> | null }

 const DETECTIVES = [
@ -60,7 +60,7 @@ async function loadSnapshot() {

  const [hyp, ev, ctr, gap, wit, jobs] = await Promise.all([
    pgQuery<RecentHypothesis>(
-      `SELECT hypothesis_id, position, posterior, confidence_band, reviewed_by
+      `SELECT hypothesis_id, position, position_pt_br, posterior, confidence_band, reviewed_by
         FROM public.hypotheses ORDER BY created_at DESC LIMIT 4`,
    ).catch(() => []),
    pgQuery<RecentEvidence>(
@ -68,14 +68,15 @@ async function loadSnapshot() {
         FROM public.evidence ORDER BY created_at DESC LIMIT 3`,
    ).catch(() => []),
    pgQuery<RecentContradiction>(
-      `SELECT contradiction_id, topic, resolution_status
+      `SELECT contradiction_id, topic, topic_pt_br, resolution_status
         FROM public.contradictions ORDER BY created_at DESC LIMIT 3`,
    ).catch(() => []),
    pgQuery<RecentGap>(
-      `SELECT gap_id, description, scope FROM public.gaps ORDER BY created_at DESC LIMIT 3`,
+      `SELECT gap_id, description, description_pt_br, scope
+         FROM public.gaps ORDER BY created_at DESC LIMIT 3`,
    ).catch(() => []),
    pgQuery<RecentWitness>(
-      `SELECT w.witness_id, e.canonical_name, w.credibility, w.verdict
+      `SELECT w.witness_id, e.canonical_name, w.credibility, w.verdict, w.verdict_pt_br
         FROM public.witnesses w
    LEFT JOIN public.entities e ON e.entity_pk = w.person_entity_pk
        ORDER BY w.created_at DESC LIMIT 3`,
@ -191,7 +192,7 @@ export async function BureauSnapshot() {
                      {h.confidence_band ?? "—"}{post !== null && ` · ${post.toFixed(2)}`}
                    </span>
                  </div>
-                  <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{h.position}</div>
+                  <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{h.position_pt_br || h.position}</div>
                  {h.reviewed_by && (
                    <div className="text-[9px] font-mono text-[#ff3344] mt-0.5">↳ reviewed by {h.reviewed_by}</div>
                  )}
@ -213,7 +214,7 @@ export async function BureauSnapshot() {
                  <span className="text-[10px] font-mono text-[#5a6678]">{c.contradiction_id}</span>
                  <span className="text-[10px] font-mono text-[#9aa6b8]">{c.resolution_status}</span>
                </div>
-                <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{c.topic}</div>
+                <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{c.topic_pt_br || c.topic}</div>
              </div>
            ))}
          </ArtefactColumn>
@ -250,7 +251,8 @@ export async function BureauSnapshot() {
          >
            {gap.map((g) => {
              const s = (g.scope ?? {}) as Record<string, unknown>;
-              const title = (s.title as string) || g.description;
+              const title = (s.title_pt_br as string) || (s.title as string)
+                || g.description_pt_br || g.description;
              return (
                <div key={g.gap_id} className="py-1.5 border-t border-[rgba(255,210,63,0.08)] first:border-t-0">
                  <div className="flex items-baseline justify-between gap-2">
@ -279,7 +281,7 @@ export async function BureauSnapshot() {
                  <span className="text-[10px] font-mono text-[#9b5de5]">{w.credibility ?? "—"}</span>
                </div>
                <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5 font-semibold">{w.canonical_name ?? "—"}</div>
-                {w.verdict && <div className="text-[11px] text-[#9aa6b8] mt-0.5 italic">{w.verdict.slice(0, 200)}</div>}
+                {(w.verdict_pt_br || w.verdict) && <div className="text-[11px] text-[#9aa6b8] mt-0.5 italic">{(w.verdict_pt_br || w.verdict || "").slice(0, 200)}</div>}
              </div>
            ))}
          </ArtefactColumn>
--- a/web/components/doc-bureau-panel.tsx
+++ b/web/components/doc-bureau-panel.tsx
@ -17,9 +17,9 @@ import Link from "next/link";
 import { pgQuery } from "@/lib/retrieval/db";

 interface EvRow { evidence_id: string; grade: string; confidence_band: string | null; source_page_id: string }
-interface HypRow { hypothesis_id: string; position: string; confidence_band: string | null; posterior: number | string | null }
-interface CtrRow { contradiction_id: string; topic: string; resolution_status: string }
-interface GapRow { gap_id: string; description: string; scope: unknown; status: string }
+interface HypRow { hypothesis_id: string; position: string; position_pt_br: string | null; confidence_band: string | null; posterior: number | string | null }
+interface CtrRow { contradiction_id: string; topic: string; topic_pt_br: string | null; resolution_status: string }
+interface GapRow { gap_id: string; description: string; description_pt_br: string | null; scope: unknown; status: string }
 interface ReportRow { slug: string; topic: string }

 const BAND_TONE: Record<string, string> = {
@ -47,7 +47,7 @@ export async function DocBureauPanel({ docId }: { docId: string }) {
  const evIds = ev.map((e) => e.evidence_id);
  const hyp: HypRow[] = evIds.length > 0
    ? await pgQuery<HypRow>(
-        `SELECT hypothesis_id, position, confidence_band, posterior
+        `SELECT hypothesis_id, position, position_pt_br, confidence_band, posterior
           FROM public.hypotheses
          WHERE EXISTS (
            SELECT 1 FROM jsonb_array_elements(evidence_refs) er
@ -60,7 +60,7 @@ export async function DocBureauPanel({ docId }: { docId: string }) {

  // Contradictions whose chunks[] has any chunk with this doc_id.
  const ctr: CtrRow[] = await pgQuery<CtrRow>(
-    `SELECT contradiction_id, topic, resolution_status
+    `SELECT contradiction_id, topic, topic_pt_br, resolution_status
       FROM public.contradictions
      WHERE EXISTS (
        SELECT 1 FROM jsonb_array_elements(chunks) c
@ -72,7 +72,7 @@ export async function DocBureauPanel({ docId }: { docId: string }) {

  // Outliers (gaps with scope.doc_id matching).
  const gap: GapRow[] = await pgQuery<GapRow>(
-    `SELECT gap_id, description, scope, status
+    `SELECT gap_id, description, description_pt_br, scope, status
       FROM public.gaps
      WHERE scope->>'doc_id' = $1
      ORDER BY gap_id LIMIT 8`,
@ -147,7 +147,7 @@ export async function DocBureauPanel({ docId }: { docId: string }) {
                      {h.confidence_band ?? "—"}{post !== null && ` · ${post.toFixed(2)}`}
                    </span>
                  </div>
-                  <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{h.position}</div>
+                  <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{h.position_pt_br || h.position}</div>
                </Link>
              );
            })}
@ -181,7 +181,7 @@ export async function DocBureauPanel({ docId }: { docId: string }) {
                  <span className="text-[10px] font-mono text-[#5a6678]">{c.contradiction_id}</span>
                  <span className="text-[10px] font-mono text-[#9aa6b8]">{c.resolution_status}</span>
                </div>
-                <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{c.topic}</div>
+                <div className="text-[12px] text-[#cbd2dd] leading-snug mt-0.5">{c.topic_pt_br || c.topic}</div>
              </div>
            ))}
          </Panel>
@ -191,7 +191,8 @@ export async function DocBureauPanel({ docId }: { docId: string }) {
          <Panel title="Outliers" color="text-[#ffd23f]" border="border-[rgba(255,210,63,0.25)]">
            {gap.map((g) => {
              const s = (g.scope ?? {}) as Record<string, unknown>;
-              const title = (s.title as string) || g.description;
+              const title = (s.title_pt_br as string) || (s.title as string)
+                || g.description_pt_br || g.description;
              return (
                <div key={g.gap_id} className="py-1.5 border-t border-[rgba(255,210,63,0.08)] first:border-t-0">
                  <div className="flex items-baseline justify-between gap-2">
--- a/web/components/job-status-poller.tsx
+++ b/web/components/job-status-poller.tsx
@ -52,9 +52,13 @@ interface EvidenceItem {
 interface HypothesisItem {
  hypothesis_id: string;
  question: string | null;
+  question_pt_br?: string | null;
  position: string | null;
+  position_pt_br?: string | null;
  argument_for: string | null;
+  argument_for_pt_br?: string | null;
  argument_against: string | null;
+  argument_against_pt_br?: string | null;
  prior: number | string | null;
  posterior: number | string | null;
  confidence_band: string | null;
@ -66,15 +70,18 @@ interface ContradictionPositionItem {
  chunk_id: string;
  page: number;
  statement: string;
+  statement_pt_br?: string | null;
  stance?: string | null;
 }

 interface ContradictionItem {
  contradiction_id: string;
  topic: string;
+  topic_pt_br?: string | null;
  chunks: ContradictionPositionItem[];
  resolution_status: string | null;
  notes: string | null;
+  notes_pt_br?: string | null;
  detected_by: string | null;
 }

@ -92,9 +99,12 @@ interface WitnessItem {
  entity_id: string | null;
  credibility: string | null;
  access_to_event: string | null;
+  access_to_event_pt_br?: string | null;
  bias_notes: string | null;
+  bias_notes_pt_br?: string | null;
  corroboration_refs: WitnessCorrItem[];
  verdict: string | null;
+  verdict_pt_br?: string | null;
 }

 interface CaseReportOutput {
@ -108,17 +118,23 @@ interface CaseReportOutput {
 interface GapItem {
  gap_id: string;
  description: string;
+  description_pt_br?: string | null;
  scope: {
    kind?: string;
    title?: string;
+    title_pt_br?: string;
    doc_id?: string;
    chunk_id?: string;
    page?: number;
    dominant_model?: string;
+    dominant_model_pt_br?: string;
    why_surprising?: string;
+    why_surprising_pt_br?: string;
    what_it_implies?: string;
+    what_it_implies_pt_br?: string;
  } | null;
  suggested_next_move: string | null;
+  suggested_next_move_pt_br?: string | null;
  status: string;
  created_by: string;
 }
@ -416,9 +432,14 @@ function HypothesisCard({ h }: { h: HypothesisItem }) {
        )}
      </div>

-      <div className="text-[14px] text-[#e7ecf3] leading-snug font-medium mb-3">
-        {h.position}
+      <div className="text-[14px] text-[#e7ecf3] leading-snug font-medium mb-1">
+        {h.position_pt_br || h.position}
      </div>
+      {h.position_pt_br && h.position && h.position_pt_br !== h.position && (
+        <div className="text-[11px] text-[#5a6678] leading-snug italic mb-3">
+          {h.position}
+        </div>
+      )}

      {(prior !== null || posterior !== null) && (
        <div className="grid grid-cols-2 gap-3 mb-3 text-[11px] font-mono">
@ -435,16 +456,28 @@ function HypothesisCard({ h }: { h: HypothesisItem }) {
        </div>
      )}

-      {h.argument_for && (
+      {(h.argument_for_pt_br || h.argument_for) && (
        <div className="mt-3">
-          <div className="text-[10px] font-mono text-[#06d6a0] uppercase mb-1">Argumento a favor</div>
-          <ArgumentBody text={h.argument_for} />
+          <div className="text-[10px] font-mono text-[#06d6a0] uppercase mb-1">Argumento a favor (PT-BR)</div>
+          <ArgumentBody text={h.argument_for_pt_br || h.argument_for!} />
+          {h.argument_for && h.argument_for_pt_br && (
+            <details className="mt-2">
+              <summary className="text-[10px] font-mono text-[#5a6678] cursor-pointer">Argument for (EN)</summary>
+              <div className="mt-1"><ArgumentBody text={h.argument_for} /></div>
+            </details>
+          )}
        </div>
      )}
-      {h.argument_against && (
+      {(h.argument_against_pt_br || h.argument_against) && (
        <div className="mt-3">
-          <div className="text-[10px] font-mono text-[#ff6ec7] uppercase mb-1">Argumento contra</div>
-          <ArgumentBody text={h.argument_against} />
+          <div className="text-[10px] font-mono text-[#ff6ec7] uppercase mb-1">Argumento contra (PT-BR)</div>
+          <ArgumentBody text={h.argument_against_pt_br || h.argument_against!} />
+          {h.argument_against && h.argument_against_pt_br && (
+            <details className="mt-2">
+              <summary className="text-[10px] font-mono text-[#5a6678] cursor-pointer">Argument against (EN)</summary>
+              <div className="mt-1"><ArgumentBody text={h.argument_against} /></div>
+            </details>
+          )}
        </div>
      )}
    </div>
@ -540,7 +573,7 @@ function GapCard({ g }: { g: GapItem }) {
        </span>
      </div>
      <div className="text-[14px] text-[#e7ecf3] font-medium mb-2">
-        {s.title || g.description}
+        {s.title_pt_br || s.title || g.description_pt_br || g.description}
      </div>
      {s.doc_id && s.chunk_id && pageStr && (
        <div className="text-[10px] font-mono mb-2">
@ -553,27 +586,27 @@ function GapCard({ g }: { g: GapItem }) {
          </Link>
        </div>
      )}
-      {s.dominant_model && (
+      {(s.dominant_model_pt_br || s.dominant_model) && (
        <div className="mt-2 p-2 bg-[#060a13] rounded border border-[rgba(255,210,63,0.08)]">
-          <div className="text-[10px] font-mono text-[#9aa6b8] uppercase mb-1">Dominant model</div>
-          <div className="text-[12px] text-[#cbd2dd] leading-relaxed">{s.dominant_model}</div>
+          <div className="text-[10px] font-mono text-[#9aa6b8] uppercase mb-1">Modelo dominante</div>
+          <div className="text-[12px] text-[#cbd2dd] leading-relaxed">{s.dominant_model_pt_br || s.dominant_model}</div>
        </div>
      )}
-      {s.why_surprising && (
+      {(s.why_surprising_pt_br || s.why_surprising) && (
        <div className="mt-2 p-2 bg-[#060a13] rounded border border-[rgba(255,210,63,0.08)]">
-          <div className="text-[10px] font-mono text-[#ffd23f] uppercase mb-1">Why surprising</div>
-          <div className="text-[12px] text-[#e7ecf3] leading-relaxed">{s.why_surprising}</div>
+          <div className="text-[10px] font-mono text-[#ffd23f] uppercase mb-1">Por que é surpreendente</div>
+          <div className="text-[12px] text-[#e7ecf3] leading-relaxed">{s.why_surprising_pt_br || s.why_surprising}</div>
        </div>
      )}
-      {s.what_it_implies && (
+      {(s.what_it_implies_pt_br || s.what_it_implies) && (
        <div className="mt-2 p-2 bg-[#060a13] rounded border border-[rgba(255,210,63,0.08)]">
-          <div className="text-[10px] font-mono text-[#ff8a4d] uppercase mb-1">What it implies</div>
-          <div className="text-[12px] text-[#cbd2dd] leading-relaxed">{s.what_it_implies}</div>
+          <div className="text-[10px] font-mono text-[#ff8a4d] uppercase mb-1">O que implica</div>
+          <div className="text-[12px] text-[#cbd2dd] leading-relaxed">{s.what_it_implies_pt_br || s.what_it_implies}</div>
        </div>
      )}
-      {g.suggested_next_move && (
+      {(g.suggested_next_move_pt_br || g.suggested_next_move) && (
        <div className="mt-2 text-[11px] font-mono text-[#06d6a0]">
-          → {g.suggested_next_move}
+          → {g.suggested_next_move_pt_br || g.suggested_next_move}
        </div>
      )}
    </div>
@ -607,23 +640,30 @@ function WitnessCard({ w }: { w: WitnessItem }) {
        )}
      </div>

-      {w.verdict && (
-        <blockquote className="text-[13px] text-[#e7ecf3] italic border-l-2 border-[#9b5de5] pl-3 my-2">
-          {w.verdict}
-        </blockquote>
+      {(w.verdict_pt_br || w.verdict) && (
+        <>
+          <blockquote className="text-[13px] text-[#e7ecf3] italic border-l-2 border-[#9b5de5] pl-3 my-2">
+            {w.verdict_pt_br || w.verdict}
+          </blockquote>
+          {w.verdict_pt_br && w.verdict && w.verdict_pt_br !== w.verdict && (
+            <blockquote className="text-[11px] text-[#5a6678] italic border-l border-[rgba(155,93,229,0.3)] pl-3 mb-2">
+              {w.verdict}
+            </blockquote>
+          )}
+        </>
      )}

      <div className="grid md:grid-cols-2 gap-3 mt-3">
-        {w.access_to_event && (
+        {(w.access_to_event_pt_br || w.access_to_event) && (
          <div className="p-2 rounded bg-[#060a13] border border-[rgba(155,93,229,0.08)]">
-            <div className="text-[10px] font-mono text-[#9b5de5] uppercase mb-1">Access to event</div>
-            <div className="text-[11px] text-[#cbd2dd] leading-relaxed">{w.access_to_event}</div>
+            <div className="text-[10px] font-mono text-[#9b5de5] uppercase mb-1">Acesso ao evento (PT-BR)</div>
+            <div className="text-[11px] text-[#cbd2dd] leading-relaxed">{w.access_to_event_pt_br || w.access_to_event}</div>
          </div>
        )}
-        {w.bias_notes && (
+        {(w.bias_notes_pt_br || w.bias_notes) && (
          <div className="p-2 rounded bg-[#060a13] border border-[rgba(155,93,229,0.08)]">
-            <div className="text-[10px] font-mono text-[#ff8a4d] uppercase mb-1">Bias notes</div>
-            <div className="text-[11px] text-[#cbd2dd] leading-relaxed">{w.bias_notes}</div>
+            <div className="text-[10px] font-mono text-[#ff8a4d] uppercase mb-1">Notas de viés (PT-BR)</div>
+            <div className="text-[11px] text-[#cbd2dd] leading-relaxed">{w.bias_notes_pt_br || w.bias_notes}</div>
          </div>
        )}
      </div>
@ -672,21 +712,30 @@ function ContradictionCard({ c }: { c: ContradictionItem }) {
          </span>
        )}
      </div>
-      <div className="text-[14px] text-[#e7ecf3] leading-snug font-medium mb-3">
-        {c.topic}
+      <div className="text-[14px] text-[#e7ecf3] leading-snug font-medium mb-1">
+        {c.topic_pt_br || c.topic}
      </div>
+      {c.topic_pt_br && c.topic_pt_br !== c.topic && (
+        <div className="text-[11px] text-[#5a6678] italic mb-3">{c.topic}</div>
+      )}

      <div className="space-y-2">
        {c.chunks.map((p, i) => {
          const pageStr = String(p.page).padStart(3, "0");
+          const stmt = p.statement_pt_br || p.statement;
          return (
            <div key={i} className="p-2 bg-[#060a13] rounded border border-[rgba(255,138,77,0.1)]">
              <div className="text-[10px] font-mono text-[#5a6678] uppercase mb-1">
                Position {i + 1}{p.stance ? ` — ${p.stance}` : ""}
              </div>
-              <blockquote className="text-[12px] text-[#e7ecf3] italic border-l-2 border-[#ff8a4d] pl-2 mb-2">
-                “{p.statement}”
+              <blockquote className="text-[12px] text-[#e7ecf3] italic border-l-2 border-[#ff8a4d] pl-2 mb-1">
+                “{stmt}”
              </blockquote>
+              {p.statement_pt_br && p.statement_pt_br !== p.statement && p.statement && (
+                <blockquote className="text-[11px] text-[#5a6678] italic border-l border-[rgba(255,138,77,0.3)] pl-2 mb-2">
+                  “{p.statement}”
+                </blockquote>
+              )}
              <Link
                href={`/d/${p.doc_id}/p${pageStr}#${p.chunk_id}`}
                className="text-[10px] font-mono text-[#7fdbff] hover:underline"
@ -698,10 +747,10 @@ function ContradictionCard({ c }: { c: ContradictionItem }) {
        })}
      </div>

-      {c.notes && (
+      {(c.notes_pt_br || c.notes) && (
        <div className="mt-3 p-2 bg-[#060a13] rounded border border-[rgba(255,138,77,0.08)]">
-          <div className="text-[10px] font-mono text-[#5a6678] uppercase mb-1">Notes</div>
-          <div className="text-[12px] text-[#cbd2dd] leading-relaxed">{c.notes}</div>
+          <div className="text-[10px] font-mono text-[#5a6678] uppercase mb-1">Notas</div>
+          <div className="text-[12px] text-[#cbd2dd] leading-relaxed">{c.notes_pt_br || c.notes}</div>
        </div>
      )}
    </div>
--- a/web/lib/i18n/pick.ts
+++ b/web/lib/i18n/pick.ts
@ -0,0 +1,36 @@
+/**
+ * pickLang — read the locale-preferred field with EN fallback.
+ *
+ * The bureau stores every narrative as a pair: `field` (EN) and
+ * `field_pt_br` (Brazilian Portuguese). UI components call this helper to
+ * surface the correct one based on the request locale.
+ *
+ *  - locale "pt-br" or "pt"  → prefer PT-BR, fall back to EN
+ *  - locale "en"              → prefer EN, fall back to PT-BR
+ *
+ * Empty / whitespace-only strings are treated as missing so a partial row
+ * still surfaces the language that has content.
+ */
+
+export type Locale = "pt-br" | "pt" | "en" | string | null | undefined;
+
+function isPt(locale: Locale): boolean {
+  return locale === "pt-br" || locale === "pt";
+}
+
+function nonEmpty(s: string | null | undefined): string | null {
+  if (typeof s !== "string") return null;
+  const t = s.trim();
+  return t.length > 0 ? s : null;
+}
+
+export function pickLang(
+  en: string | null | undefined,
+  pt_br: string | null | undefined,
+  locale: Locale,
+): string | null {
+  const enValid = nonEmpty(en);
+  const ptValid = nonEmpty(pt_br);
+  if (isPt(locale)) return ptValid ?? enValid;
+  return enValid ?? ptValid;
+}