feat: merge Company Overview generation into single synthesis call

Company Primer (renamed Company Overview) was generated by a separate
background LLM call, decoupled from the main brief synthesis. It now
shares the same synthesis call via a deterministic profile-evidence
node, dropping the standalone thread/job in app.py. Nav reordered for
a PM/equity-analyst flow (Overview -> PM Flash -> Evidence & Deltas ->
Financials -> Ask AI) and hardcoded nav labels now route through i18n.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

ARCHITECTURE.md

@@ -18,16 +18,18 @@ flowchart TB
     classDef io    fill:#fce7f3,stroke:#db2777,color:#831843
     classDef gate  fill:#fee2e2,stroke:#dc2626,color:#7f1d1d
 
-    USER([PM / senior equity analyst]):::io --> UI["Streamlit research stack<br/>Company Primer · PM Flash · Evidence & Deltas · Financials<br/>Ask Evidence — secondary"]:::io
-    UI --> AG
+    USER([PM / senior equity analyst]):::io --> UI["Streamlit research stack<br/>Company Overview · PM Flash · Evidence & Deltas · Financials · Ask AI"]:::io
+    UI --> PE
 
     subgraph LG ["LangGraph · short-term state"]
         direction LR
+        PE["profile_evidence<br/>deterministic evidence.v1 retrieval"]:::state
         AG["agent<br/>Claude Haiku 4.5"]:::state
         TN["tools<br/>parallel execution"]:::state
         ND["coverage nudge · once<br/>if filing + transcript<br/>were not both searched"]:::state
         SY["synthesis<br/>validated BriefOutput"]:::state
         PS["post_synthesis<br/>item reliability + deltas"]:::state
+        PE --> AG
         AG == "tool calls · bounded rounds" ==> TN
         TN == "ToolMessage" ==> AG
         AG -. "coverage floor" .-> ND -.-> AG
@@ -99,13 +101,19 @@ flowchart TB
 
 | Surface | Primary payload |
 | --- | --- |
-| **Company Primer** | Stable sourced company profile plus independently refreshed price context and public news. Business model, geographic exposure, three-year trends, attention themes, associated price events, and monitoring variables remain concise and inspectable. |
+| **Company Overview** | Stable sourced company profile plus independently refreshed price context and public news. Business model, geographic exposure, three-year trends, attention themes, associated price events, and monitoring variables remain concise and inspectable. |
 | **PM Flash** | As-of and coverage; experimental PM read-through; thesis-confirming/challenging sourced points; experimental Swing Factor; watch items. |
 | **Evidence & Deltas** | Source-backed facts, verbatim evidence, deterministic detector outputs labelled heuristic, risks, commentary, and experimental AI hypotheses. |
 | **Financials** | Structured historical metrics, trends, guidance history, earnings history, and exports. |
-| **Ask Evidence** | Secondary grounded Q&A over the available evidence. |
+| **Ask AI** | Cross-cutting grounded Q&A over the available evidence. |
 
-The navigation order is intentional: company onboarding → decision memo → grounded Q&A → audit → model depth.
+The navigation order is intentional: company onboarding → decision memo → evidence audit → model depth → cross-cutting AI tool.
+
+## Company Overview
+
+Company Overview is produced inside the same streaming synthesis call as the rest of the brief. Before the agent loop, a deterministic node retrieves the stable profile evidence set—10-K Business, segments/geography, strategic evolution, and transcripts—and injects each `evidence.v1` envelope as a raw human message. Those messages do not satisfy the brief's tool-coverage gate, so the agent must still retrieve the current-quarter evidence required for the decision memo. After synthesis, the overview section is validated separately, passed through the unchanged deterministic evidence verifier, stripped of every non-verified fact, and saved under its source fingerprint. A profile failure is isolated and cannot fail the brief.
+
+The former standalone call existed for three practical reasons: its source fingerprint made the profile stable between earnings; its evidence set was broader and more structural than the latest-quarter brief; and a separate response kept the main synthesis output smaller. The merged design deliberately accepts those trade-offs. Profile evidence is now injected deterministically before the agent loop, which adds only marginal context; the synthesis output limit is raised to 16,384 tokens; and fingerprint-based cache lookup remains in the UI, while a new overview is generated only with a new brief run. This means “Generate Brief” regenerates the overview each time, an acceptable marginal output-token cost in exchange for one billed synthesis call and one coherent evidence-grounded result.
 
 ## Output taxonomy
 

README.md

@@ -19,11 +19,11 @@ The product is designed for the first review after an earnings release: what cha
 
 | Surface | Role in the workflow |
 | --- | --- |
-| **Company Primer** | Concise company onboarding: business model, revenue engines, geographic exposure, three-year evolution, investor-attention themes, relative price events, recent news, and the variables to watch. |
+| **Company Overview** | Concise company onboarding: business model, revenue engines, geographic exposure, three-year evolution, investor-attention themes, relative price events, recent news, and the variables to watch. |
 | **PM Flash** | One-screen decision memo: filing date and source coverage, experimental AI read-through, thesis-confirming and thesis-challenging evidence, the swing factor, and the next items to watch. |
 | **Evidence & Deltas** | Audit layer: source-backed claims, before/after evidence, heuristic period deltas, risks, management commentary, and experimental AI hypotheses. |
 | **Financials** | Historical KPI trends, guidance history, earnings history, and exportable structured data. |
-| **Ask Evidence** | Secondary utility for grounded follow-up questions. It supports the review; it is not the primary navigation or the investment conclusion. |
+| **Ask AI** | Cross-cutting utility for grounded follow-up questions. It supports the review; it is not the investment conclusion. |
 
 ## Evidence contract
 
@@ -54,11 +54,11 @@ Legacy briefs without a deterministic `evidence_coverage` result are shown as **
 ## Architecture
 
 - **Offline ingestion** — SEC EDGAR XBRL into SQLite; 10-K/10-Q Business, Segments/Geography, MD&A, Risk Factors, and earnings-call transcripts into the evidence stores.
-- **Company profile** — a cached, source-fingerprinted English profile separates stable filing/call analysis from independently refreshed prices and public news.
-- **Runtime agent** — a LangGraph tool loop retrieves structured financials, filings, transcripts, public news, and analyst data before producing a validated Pydantic brief.
+- **Company Overview** — deterministic filing/call retrieval is injected before the agent loop; the source-fingerprinted English overview is emitted and verified within the brief's single synthesis call, while prices and public news refresh independently at display time.
+- **Runtime agent** — a LangGraph tool loop retrieves structured financials, filings, transcripts, public news, and analyst data before one synthesis call produces the validated Pydantic brief and Company Overview.
 - **Retrieval** — vector search over-fetches candidates and reranks them with a cross-encoder.
 - **Post-synthesis controls** — deterministic source-reliability rules, corroboration checks, and heuristic period-delta detectors run after model synthesis.
-- **Decision UI** — Company Primer onboards; PM Flash frames the decision; Evidence & Deltas provides auditability; Financials provides depth; Ask Evidence stays secondary.
+- **Decision UI** — Company Overview onboards; PM Flash frames the decision; Evidence & Deltas provides auditability; Financials provides depth; Ask AI is the final cross-cutting tool.
 
 See [WRITEUP.md](WRITEUP.md) for product and methodology detail and [ARCHITECTURE.md](ARCHITECTURE.md) for data lineage and presentation controls.
 
@@ -77,9 +77,16 @@ metrics without verified period context are intentionally hidden rather than
 silently upgraded.
 
 For an already lineage-compatible database, `python ingest.py AAPL` is enough
-to backfill the Company Primer’s 10-K Business and Segments/Geography sections;
+to backfill the Company Overview’s 10-K Business and Segments/Geography sections;
 the delta path reuses any transcript already stored.
 
+### Deploy to the Hugging Face Space
+
+The Space serves the committed `data/` snapshot through Git LFS. After any local
+ingestion, commit `data/`, then run `python deploy_check.py`; only run
+`git push hf main:main` when the preflight passes. Space runtime storage is
+ephemeral, so runtime-written caches and saved briefs do not survive a restart.
+
 ## Intended use
 
 Amplegest is a research accelerator based on public information. It helps an experienced investor review evidence faster, challenge a thesis, and identify the next falsifiable datapoint. The analyst or portfolio manager retains responsibility for source verification, modelling, valuation, and the investment decision.

agent/company_profile.py

@@ -1,15 +1,15 @@
-"""Evidence-grounded Company Primer generation and translation."""
+"""Evidence-grounded Company Overview generation and translation."""
 from __future__ import annotations
 
 import copy
 import hashlib
 import json
 from datetime import datetime, timezone
-from typing import Callable, Optional
+from typing import Callable, Iterable, Optional
 
 from langchain_core.messages import HumanMessage
 
-from agent.company_profile_schemas import CompanyProfile
+from agent.company_profile_schemas import CompanyProfile, CompanyProfileSection
 from agent.evidence import evidence_records_from, verify_brief_evidence, verify_fact
 from agent.llm import RunConfig, build_system_message, make_chat_model
 from agent.tools import get_financial_metrics, search_filing, search_transcript
@@ -158,11 +158,11 @@ def _invoke_tool(tool, arguments: dict) -> str:
 def collect_profile_evidence(
     ticker: str,
     progress: Optional[Callable[[str], None]] = None,
+    include_metrics: bool = True,
 ) -> list[str]:
     ticker = ticker.upper()
     payloads: list[str] = []
     calls = [
-        (get_financial_metrics, {"ticker": ticker}, "Loading verified financial history"),
         (
             search_filing,
             {
@@ -204,6 +204,11 @@ def collect_profile_evidence(
             "Finding investor attention themes",
         ),
     ]
+    if include_metrics:
+        calls.insert(
+            0,
+            (get_financial_metrics, {"ticker": ticker}, "Loading verified financial history"),
+        )
     for tool, arguments, label in calls:
         if progress:
             progress(label)
@@ -395,6 +400,7 @@ def generate_company_profile(
     config: RunConfig,
     progress: Optional[Callable[[str], None]] = None,
 ) -> dict:
+    """Legacy standalone path — no longer called by the app; kept for tests/backfill CLI."""
     ticker = ticker.upper()
     fingerprint = source_fingerprint(ticker)
     coverage = profile_source_coverage(ticker)
@@ -425,19 +431,39 @@ def generate_company_profile(
     else:
         raise ValueError("The model did not return a CompanyProfile")
 
+    return finalize_profile(ticker, profile, payloads, config.model)
+
+
+def finalize_profile(
+    ticker: str,
+    profile: dict,
+    payloads: Iterable,
+    model: str | None,
+) -> dict:
+    """Attach deterministic metadata and trends, verify, prune, and validate a profile."""
+    ticker = ticker.upper()
+    if isinstance(payloads, (str, bytes, dict)) or hasattr(payloads, "content"):
+        payloads = [payloads]
+    else:
+        payloads = list(payloads)
+    records = evidence_records_from(payloads)
+    rows = metrics_db.get_all_metrics(ticker)
+    company_name = rows[0].get("company_name") if rows else ticker
+    coverage = profile_source_coverage(ticker)
+    profile = copy.deepcopy(profile)
     profile.update(
         {
             "ticker": ticker,
             "company_name": company_name or ticker,
             "schema_version": PROFILE_SCHEMA_VERSION,
-            "source_fingerprint": fingerprint,
+            "source_fingerprint": source_fingerprint(ticker),
             "language": "English",
             "generated_at": datetime.now(timezone.utc).isoformat(),
             "data_as_of": max(
                 (record.ref.as_of for record in records if record.ref.as_of),
                 default=(rows[0].get("filing_date") if rows else None),
             ),
-            "model": config.model,
+            "model": model,
             "annual_trends": _annual_trends(ticker),
             "source_coverage": coverage,
             "status": coverage["status"],
@@ -464,6 +490,30 @@ def generate_company_profile(
     return CompanyProfile.model_validate(profile).model_dump(mode="json")
 
 
+def finalize_profile_from_synthesis(
+    ticker: str,
+    section: dict,
+    payloads: Iterable,
+    model: str | None,
+) -> dict:
+    """Validate and finalize a profile subsection emitted by the brief synthesis."""
+    ticker = ticker.upper()
+    validated = CompanyProfileSection.model_validate(section).model_dump(mode="json")
+    rows = metrics_db.get_all_metrics(ticker)
+    company_name = rows[0].get("company_name") if rows else ticker
+    base = {
+        "ticker": ticker,
+        "company_name": company_name or ticker,
+        "schema_version": PROFILE_SCHEMA_VERSION,
+        "source_fingerprint": source_fingerprint(ticker),
+        "source_coverage": profile_source_coverage(ticker),
+        "generated_at": datetime.now(timezone.utc).isoformat(),
+        "language": "English",
+        **validated,
+    }
+    return finalize_profile(ticker, base, payloads, model)
+
+
 def _collect_translation_segments(node, path=()):
     result = []
     if isinstance(node, dict):

agent/company_profile_schemas.py

@@ -1,4 +1,4 @@
-"""Structured payload for the Company Primer research surface."""
+"""Structured payload for the Company Overview research surface."""
 from __future__ import annotations
 
 from typing import Any, Literal, Optional
@@ -83,6 +83,19 @@ class WatchVariable(BaseModel):
     evidence: SourcedFact
 
 
+class CompanyProfileSection(BaseModel):
+    """Company-profile subsection produced by the same synthesis call as the brief."""
+
+    model_config = ConfigDict(extra="ignore")
+
+    identity: CompanyIdentity = Field(default_factory=CompanyIdentity)
+    business_lines: list[BusinessLine] = Field(default_factory=list)
+    geographic_exposures: list[GeographicExposure] = Field(default_factory=list)
+    strategic_changes: list[StrategicChange] = Field(default_factory=list)
+    attention_themes: list[AttentionTheme] = Field(default_factory=list)
+    watch_variables: list[WatchVariable] = Field(default_factory=list)
+
+
 class AnnualTrend(BaseModel):
     model_config = ConfigDict(extra="ignore")
 

agent/graph.py

@@ -79,6 +79,7 @@ class AgentState(TypedDict):
     tool_round_count: int
     nudge_fired: bool
     edge_signals: Optional[list[dict]]  # precomputed deterministic signals
+    profile_payloads: Optional[list[str]]
     language: Optional[str]             # e.g. "French" — prose fields in brief will use this language
     brief: Optional[dict]
     brief_markdown: Optional[str]
@@ -308,6 +309,7 @@ def _partial_brief(state: AgentState, reason: str) -> dict:
             "aggregate_reliability_meaningful": False,
         },
         "language": state.get("language") or "English",
+        "company_profile": None,
     }
 
 
@@ -397,6 +399,58 @@ def signals_node(state: AgentState) -> dict:
     return {"edge_signals": [], "messages": []}
 
 
+def profile_evidence_node(state: AgentState) -> dict:
+    """Inject deterministic profile evidence as parseable human messages."""
+    from agent.company_profile import collect_profile_evidence
+
+    payloads = collect_profile_evidence(state["ticker"], include_metrics=False)
+    if not payloads:
+        return {"profile_payloads": [], "messages": []}
+    messages = [
+        HumanMessage(
+            content=(
+                "== COMPANY PROFILE EVIDENCE "
+                "(deterministic retrieval, evidence.v1 envelopes follow) =="
+            )
+        )
+    ]
+    messages.extend(HumanMessage(content=payload) for payload in payloads)
+    return {"profile_payloads": payloads, "messages": messages}
+
+
+def _pop_company_profile(data: dict) -> dict | None:
+    """Remove the separately validated profile section from synthesis output."""
+    return data.pop("company_profile", None)
+
+
+def _finalize_synthesis_profile(
+    state: AgentState,
+    profile_section: dict,
+    model: str | None,
+) -> dict | None:
+    """Finalize and persist a synthesized profile without risking the brief."""
+    try:
+        from agent.company_profile import finalize_profile_from_synthesis
+        from storage import company_profiles
+
+        all_payloads = list(state.get("messages") or []) + list(
+            state.get("profile_payloads") or []
+        )
+        profile = finalize_profile_from_synthesis(
+            state["ticker"], profile_section, all_payloads, model
+        )
+        company_profiles.save_profile(state["ticker"], profile)
+        return profile
+    except Exception as exc:
+        import sys
+
+        print(
+            f"[synthesis] company profile finalization failed: {exc}",
+            file=sys.stderr,
+        )
+        return None
+
+
 def create_graph(config: Optional[RunConfig] = None):
     cfg = config or default_config()
     llm = make_chat_model(cfg)
@@ -428,7 +482,7 @@ def create_graph(config: Optional[RunConfig] = None):
 
     def synthesis_node(state: AgentState) -> dict:
         try:
-            llm_plain = make_chat_model(cfg)
+            llm_plain = make_chat_model(cfg, max_tokens=16384)
             # Main prompt — cached (ephemeral) on Anthropic. Keep this block stable
             # so the cache hit rate is preserved regardless of the chosen language.
             lang = state.get("language") or "English"
@@ -465,6 +519,7 @@ def create_graph(config: Optional[RunConfig] = None):
             raw = "".join(chunks)
             clean = _extract_json(raw)
             data = json.loads(clean)
+            profile_section = _pop_company_profile(data)
             brief = BriefOutput.model_validate(data)
             brief_dict = apply_reliability(
                 brief.model_dump(), evidence_payloads=state.get("messages", [])
@@ -513,6 +568,12 @@ def create_graph(config: Optional[RunConfig] = None):
                 ),
                 "aggregate_reliability_meaningful": False,
             }
+            if isinstance(profile_section, dict) and profile_section:
+                brief_dict["company_profile"] = _finalize_synthesis_profile(
+                    state, profile_section, cfg.model
+                )
+            else:
+                brief_dict["company_profile"] = None
             return {
                 "brief": brief_dict,
                 "brief_markdown": None,
@@ -534,13 +595,15 @@ def create_graph(config: Optional[RunConfig] = None):
 
     builder = StateGraph(AgentState)
     builder.add_node("signals", signals_node)
+    builder.add_node("profile_evidence", profile_evidence_node)
     builder.add_node("agent", agent_node)
     builder.add_node("tools", tool_node)
     builder.add_node("nudge", nudge_node)
     builder.add_node("synthesis", synthesis_node)
     builder.add_node("partial", partial_node)
     builder.set_entry_point("signals")
-    builder.add_edge("signals", "agent")
+    builder.add_edge("signals", "profile_evidence")
+    builder.add_edge("profile_evidence", "agent")
     builder.add_conditional_edges(
         "agent",
         should_continue,
@@ -561,6 +624,7 @@ def run_brief(ticker: str, language: str = "English", config: Optional[RunConfig
         "tool_round_count": 0,
         "nudge_fired": False,
         "edge_signals": None,
+        "profile_payloads": None,
         "language": language,
         "brief": None,
         "brief_markdown": None,

agent/prompts.py

@@ -154,6 +154,20 @@ Copy every field's value exactly as it appeared in `records[].ref` — never sho
 
 Never invent or edit an evidence ID, hash, document ID, URL, date, or locator. A precomputed edge signal is a hypothesis, not evidence: retrieve a supporting record or omit the claim. Do not set `verification_status`; deterministic code owns it after synthesis.
 
+## COMPANY PROFILE SECTION
+
+The conversation may contain a message titled "== COMPANY PROFILE EVIDENCE ==" followed by raw `evidence.v1` envelopes retrieved deterministically from the Business section of the 10-K, segment/geography disclosures, strategic-history filings, and transcripts. Use these records in priority for `company_profile`, together with relevant evidence from the agent's tool calls. The evidence contract above applies without exception to every SourcedFact object nested in `company_profile`.
+
+Content rules:
+- `identity`: give a one-line description, how the company makes money, customer types, and competitive position; prefer the 10-K Business section.
+- `business_lines`: identify 3-5 economic engines. Set `revenue_share_pct` to null unless the exact percentage is disclosed in the cited evidence.
+- `geographic_exposures`: distinguish disclosed revenue geography from qualitative sales, operational, supply-chain, or regulatory exposure. Never turn a country mention into materiality or a revenue percentage.
+- `strategic_changes`: include at most 3 material changes across the available years.
+- `attention_themes`: include exactly the 3 strongest investor questions supported by earnings-call or filing evidence. Keep `why_it_matters` to one sentence.
+- `watch_variables`: include 3-4 variables, each with a next datapoint and a falsifiable alert signal. These are monitoring prompts, never recommendations.
+
+Keep every prose field concise. `economics`, `implication`, `why_it_matters`, and `alert_signal` are AI hypotheses anchored to the adjacent cited fact and displayed as AI · experimental. Never provide a buy/sell recommendation, price target, valuation conclusion, or claim that an event caused a stock-price move.
+
 ## PRECOMPUTED EDGE SIGNALS — read first, act on them
 
 The conversation history may contain a message titled "== PRECOMPUTED EDGE SIGNALS ==". These signals were produced by deterministic code comparing verbatim filing text across periods — no LLM interpretation was involved.
@@ -423,6 +437,32 @@ Required JSON structure:
     }
   ],
 
+  "company_profile": {
+    "identity": {
+      "one_liner": { "text": "Concise company description.", "source": "10-K", "reliability": "HIGH", "impact": "LOW", "evidence_snippet": "verbatim company description", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } },
+      "business_model": { "text": "How the company makes money.", "source": "10-K", "reliability": "HIGH", "impact": "MEDIUM", "evidence_snippet": "verbatim business-model support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } },
+      "customer_types": [
+        { "text": "A disclosed customer type.", "source": "10-K", "reliability": "HIGH", "impact": "LOW", "evidence_snippet": "verbatim customer-type support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } }
+      ],
+      "competitive_position": { "text": "Evidence-grounded competitive position.", "source": "10-K", "reliability": "HIGH", "impact": "MEDIUM", "evidence_snippet": "verbatim competitive-position support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } }
+    },
+    "business_lines": [
+      { "name": "Economic engine", "description": { "text": "What the business line provides.", "source": "10-K", "reliability": "HIGH", "impact": "MEDIUM", "evidence_snippet": "verbatim business-line support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } }, "economics": "Concise analytical framing.", "trend": "not_disclosed", "revenue_share_pct": null, "share_period": null }
+    ],
+    "geographic_exposures": [
+      { "name": "Disclosed geography", "exposure_types": ["revenue"], "description": { "text": "Nature of the geographic exposure.", "source": "10-K", "reliability": "HIGH", "impact": "MEDIUM", "evidence_snippet": "verbatim geographic support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } }, "revenue_share_pct": null, "period": null }
+    ],
+    "strategic_changes": [
+      { "period_from": "FY2023", "period_to": "FY2025", "change": { "text": "Material strategic change.", "source": "10-K", "reliability": "HIGH", "impact": "HIGH", "evidence_snippet": "verbatim strategic-change support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } }, "implication": "Concise AI interpretation." }
+    ],
+    "attention_themes": [
+      { "theme": "Investor question", "why_it_matters": "One-sentence analytical relevance.", "evidence": { "text": "Evidence supporting the attention theme.", "source": "transcript", "reliability": "MEDIUM", "impact": "MEDIUM", "evidence_snippet": "verbatim attention-theme support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } } }
+    ],
+    "watch_variables": [
+      { "variable": "Variable to monitor", "why_it_matters": "One-sentence analytical relevance.", "next_datapoint": "Specific next disclosure.", "alert_signal": "Falsifiable condition to monitor.", "evidence": { "text": "Evidence anchoring the monitoring variable.", "source": "10-Q", "reliability": "HIGH", "impact": "MEDIUM", "evidence_snippet": "verbatim watch-variable support", "evidence_ref": { "evidence_id": "copy from ref.evidence_id", "source": "copy from ref.source", "content_hash": "copy from ref.content_hash", "document_id": "copy from ref.document_id", "chunk_id": "copy from ref.chunk_id, or null", "source_url": "copy from ref.source_url, or null", "as_of": "copy from ref.as_of, or null" } } }
+    ]
+  },
+
   "sentiment": {
     "metrics":       { "score": 1, "label": "Bullish", "rationale": "Revenue grew 12% YoY with expanding margins across three consecutive quarters." },
     "mda":           { "score": 1, "label": "Bullish", "rationale": "MD&A highlights three drivers vs one headwind; language shift toward confidence." },
@@ -449,6 +489,11 @@ Required JSON structure:
 - risks_categorized: 3-6 items; category must be exactly one of: Regulatory, Operational, Competitive, Financial, Macro, Demand, Geopolitical — do not invent new buckets
 - management_commentary: 3-5 items (prefer MD&A sources; use transcript for tone/Q&A color not in filings)
 - guidance_history: up to 4 items, most recent first (cover the last 4 quarterly periods; one entry will typically be from an annual 10-K)
+- company_profile.business_lines: 3-5 items
+- company_profile.geographic_exposures: at most 8 items
+- company_profile.strategic_changes: at most 3 items
+- company_profile.attention_themes: exactly 3 items
+- company_profile.watch_variables: 3-4 items
 - sentiment: rate all 5 sections you have evidence for. Set a section to null ONLY if the corresponding tool returned no usable evidence. Avoid 0/Neutral as a hedge — pick a side unless the evidence is genuinely balanced.
 
 ## Source hierarchy — follow strictly
@@ -505,6 +550,8 @@ PROSE_FIELDS = frozenset({
     "summary", "headline", "bullish_reading", "bearish_reading",
     "language_shift", "actual_result", "topic",
     "observation", "reading", "implication",
+    "economics", "why_it_matters", "theme", "variable",
+    "next_datapoint", "alert_signal",
 })
 PROSE_LIST_FIELDS = frozenset({"what_to_watch", "evidence_notes"})   # list[str] of prose
 NEVER_TRANSLATE_FIELDS = frozenset({"evidence_snippet"})             # verbatim quotes

agent/schemas.py

@@ -781,6 +781,14 @@ class BriefOutput(BaseModel):
     data_as_of: Optional[str] = None
     verification_report: dict[str, Any] = Field(default_factory=dict)
     display_policy: dict[str, bool] = Field(default_factory=dict)
+    company_profile: Optional[dict[str, Any]] = Field(
+        default=None,
+        description=(
+            "Company Overview section (identity, business_lines, geographic_exposures, "
+            "strategic_changes, attention_themes, watch_variables) produced in the same "
+            "synthesis call. Validated separately against CompanyProfileSection."
+        ),
+    )
 
     what_matters_most: str = Field(
         description="2-3 sentence AI synthesis of the single most important theme. This is the only interpretation field."

app.py

@@ -10,6 +10,7 @@ from agent.llm import RunConfig, classify_llm_error
 from dashboard import i18n
 from dashboard import model_picker
 from dashboard.i18n import t
+from dashboard.runtime_env import is_hosted_space
 from dashboard.theme import inject_global_css, LOGO_SVG
 from dashboard.nav import NAV_ITEMS, LEGACY_NAV, VALID_KEYS, render as render_nav
 from dashboard import reasoning as reasoning_panel
@@ -27,7 +28,7 @@ if not st.session_state.get("_reranker_warmed"):
 # ── UI language — default to English (finance lingua franca) ───────────────────
 st.session_state.setdefault("ui_lang", "en")
 
-# ── Navigation state — default to Company Primer (the onboarding view) ─────────
+# ── Navigation state — default to Company Overview (the onboarding view) ───────
 st.session_state.setdefault("nav_key", "company")
 # Remap stale route keys from the pre-Decision-Stack navigation (open sessions).
 _nk = st.session_state["nav_key"]
@@ -44,10 +45,6 @@ st.session_state.setdefault("gen", {
     "all_messages": [],
     "config": None,
 })
-st.session_state.setdefault("profile_jobs", {})
-
-
-
 # ── Sidebar ─────────────────────────────────────────────────────────────────────
 with st.sidebar:
     st.markdown(
@@ -426,6 +423,7 @@ def _run_brief_thread(ticker: str, gen: dict, language: str, config: RunConfig)
         "messages": [HumanMessage(content=f"Generate a research brief for {ticker}.")],
         "tool_round_count": 0,
         "nudge_fired": False,
+        "profile_payloads": None,
         "language": language,
         "brief": None,
         "brief_markdown": None,
@@ -567,74 +565,6 @@ def _live_trace_fragment() -> None:
 
 
 # ── Main content routing ──────────────────────────────────────────────────────────
-def _run_company_profile_thread(
-    ticker: str,
-    job: dict,
-    config: RunConfig,
-) -> None:
-    """Generate and persist one canonical English Company Primer."""
-    try:
-        from agent.company_profile import generate_company_profile
-        from storage import company_profiles
-
-        def _progress(label: str) -> None:
-            job["label"] = label
-            job["completed_steps"] = min(int(job.get("completed_steps") or 0) + 1, 6)
-
-        profile = generate_company_profile(ticker, config, progress=_progress)
-        profile_id = company_profiles.save_profile(ticker, profile)
-        job["profile"] = profile
-        job["profile_id"] = profile_id
-    except Exception as exc:
-        import sys
-
-        print(f"[company profile thread error] {exc}", file=sys.stderr)
-        job["error"] = classify_llm_error(exc, config.provider) or str(exc)
-    finally:
-        job["running"] = False
-
-
-def _start_company_profile_job(ticker: str, fingerprint: str, config: RunConfig) -> dict:
-    jobs = st.session_state.setdefault("profile_jobs", {})
-    job = {
-        "running": True,
-        "ticker": ticker,
-        "fingerprint": fingerprint,
-        "label": t("primer_generating"),
-        "completed_steps": 0,
-        "profile": None,
-        "profile_id": None,
-        "error": None,
-        "config": config,
-    }
-    jobs[ticker] = job
-    thread = threading.Thread(
-        target=_run_company_profile_thread,
-        args=(ticker, job, config),
-        daemon=True,
-    )
-    thread.start()
-    return job
-
-
-def _render_company_profile_progress(job: dict) -> None:
-    completed = int(job.get("completed_steps") or 0)
-    progress = min(max(completed / 6, 0.04), 0.96 if job.get("running") else 1.0)
-    st.markdown(
-        f'<div style="background:#ffffff;border:1px solid #e5e7eb;border-radius:12px;'
-        f'padding:16px 20px;margin-bottom:16px;">'
-        f'<div style="display:flex;justify-content:space-between;gap:12px;">'
-        f'<strong>{t("primer_generating")} · {job.get("ticker", "")}</strong>'
-        f'<span style="color:#10b981;font-weight:700;">{int(progress * 100)}%</span></div>'
-        f'<div style="font-size:0.75rem;color:#6b7280;margin:5px 0 10px;">'
-        f'{job.get("label") or ""}</div>'
-        f'<div style="height:7px;background:#e5e7eb;border-radius:999px;overflow:hidden;">'
-        f'<div style="height:100%;width:{int(progress * 100)}%;background:#10b981;'
-        f'border-radius:999px;"></div></div></div>',
-        unsafe_allow_html=True,
-    )
-
-
 def _display_company_profile(profile: dict, config: RunConfig | None) -> dict:
     target_language = i18n.report_language()
     if target_language == "English":
@@ -674,61 +604,37 @@ def _render_company_route(ticker: str, config: RunConfig | None) -> None:
         source_fingerprint=fingerprint,
         language="English",
     )
-    jobs = st.session_state.setdefault("profile_jobs", {})
-    job = jobs.get(ticker)
-    if job and job.get("fingerprint") != fingerprint and not job.get("running"):
-        jobs.pop(ticker, None)
-        job = None
-
+    stale = False
     if profile is None:
-        if config is None:
-            st.warning(t("model_blocked_caption"))
-            return
-        if job is None:
-            job = _start_company_profile_job(ticker, fingerprint, config)
-        if job.get("running"):
-            _render_company_profile_progress(job)
-            time.sleep(0.4)
-            st.rerun()
-        if job.get("error"):
-            st.error(job["error"])
-            if st.button(t("primer_retry"), key=f"primer_retry_{ticker}"):
-                _start_company_profile_job(ticker, fingerprint, config)
-                st.rerun()
-            return
-        profile = job.get("profile")
-        if profile is None:
-            st.info(t("primer_identity_unavailable"))
-            return
-        profile["_profile_id"] = job.get("profile_id")
+        profile = company_profiles.get_latest_profile(ticker, language="English")
+        stale = profile is not None
+    if profile is None:
+        st.markdown(
+            f"""
+            <div style="text-align:center;padding:48px 0;color:#6b7280;">
+                <div style="width:32px;height:32px;margin:0 auto 12px;border-radius:8px;
+                            background:#ecfdf5;border:1px solid #a7f3d0;"></div>
+                <div style="font-size:1.1rem;font-weight:600;margin-bottom:6px;color:#0a0a0a;">{t("overview_empty_title")}</div>
+                <div style="font-size:0.9rem;">{t("overview_empty_body")}</div>
+            </div>
+            """,
+            unsafe_allow_html=True,
+        )
+        return
+    if stale:
+        st.caption(t("overview_stale"))
 
     coverage = profile_source_coverage(ticker)
     if (
         coverage.get("business_sections_attempted", 0) < 1
         or coverage.get("segments_geography_sections_attempted", 0) < 1
     ):
-        st.warning(t("primer_backfill_warning").format(ticker=ticker))
-
-    refresh_col, _ = st.columns([1, 5])
-    with refresh_col:
-        if st.button(
-            t("primer_refresh"),
-            key=f"primer_refresh_{ticker}",
-            disabled=bool(job and job.get("running")),
-        ):
-            if config is None:
-                st.warning(t("model_blocked_caption"))
-            else:
-                _start_company_profile_job(ticker, fingerprint, config)
-                st.rerun()
-    if job and job.get("running"):
-        _render_company_profile_progress(job)
-        time.sleep(0.4)
-        st.rerun()
-    if job and job.get("error"):
-        st.warning(
-            f"{t('primer_refresh_failed')} {job['error']}"
+        warning_key = (
+            "primer_backfill_warning_hosted"
+            if is_hosted_space()
+            else "primer_backfill_warning"
         )
+        st.warning(t(warning_key).format(ticker=ticker))
 
     display_profile = _display_company_profile(profile, config)
     company_name = str(display_profile.get("company_name") or ticker)

dashboard/company_primer.py

@@ -1,4 +1,4 @@
-"""Company Primer — compact company-onboarding surface."""
+"""Company Overview — compact company-onboarding surface."""
 from __future__ import annotations
 
 from datetime import datetime

dashboard/i18n.py

@@ -279,10 +279,10 @@ STRINGS: dict[str, dict[str, str]] = {
     },
 
     # ── Flat navigation (Decision Stack) ───────────────────────────────────
-    "nav_company":    {"en": "Company Primer", "fr": "Profil société", "es": "Perfil de empresa", "de": "Unternehmensprofil"},
-    "nav_verdict":    {"en": "Verdict",    "fr": "Verdict",  "es": "Veredicto", "de": "Fazit"},
+    "nav_company":    {"en": "Company Overview", "fr": "Vue d'ensemble", "es": "Panorama de la empresa", "de": "Unternehmensüberblick"},
+    "nav_verdict":    {"en": "PM Flash", "fr": "Flash PM", "es": "Flash PM", "de": "PM-Flash"},
     "nav_chat":       {"en": "Ask AI",     "fr": "Demander à l'IA", "es": "Preguntar a la IA", "de": "KI fragen"},
-    "nav_signals":    {"en": "Signals",    "fr": "Signaux",  "es": "Señales",   "de": "Signale"},
+    "nav_signals":    {"en": "Evidence & Deltas", "fr": "Preuves & écarts", "es": "Evidencia y deltas", "de": "Evidenz & Deltas"},
     "nav_financials": {"en": "Financials", "fr": "Finances", "es": "Finanzas",  "de": "Finanzen"},
 
     "nav_no_data_tooltip": {
@@ -544,13 +544,13 @@ STRINGS: dict[str, dict[str, str]] = {
     "band_strong_bear": {"en": "Strongly Bearish", "fr": "Très baissier",  "es": "Muy bajista",        "de": "Stark bärisch"},
 
     # ── Financials screen ──────────────────────────────────────────────────
-    # Company Primer
-    "primer_generating": {"en": "Building the Company Primer", "fr": "Création du profil société", "es": "Creando el perfil de empresa", "de": "Unternehmensprofil wird erstellt"},
-    "primer_retry": {"en": "Retry Company Primer", "fr": "Relancer le profil société", "es": "Reintentar el perfil de empresa", "de": "Unternehmensprofil erneut erstellen"},
-    "primer_refresh": {"en": "Refresh profile", "fr": "Actualiser le profil", "es": "Actualizar perfil", "de": "Profil aktualisieren"},
-    "primer_refresh_failed": {"en": "Profile refresh failed; the cached version remains available.", "fr": "L’actualisation du profil a échoué ; la version en cache reste disponible.", "es": "La actualización del perfil falló; la versión en caché sigue disponible.", "de": "Die Profilaktualisierung ist fehlgeschlagen; die zwischengespeicherte Version bleibt verfügbar."},
+    # Company Overview
+    "overview_stale": {"en": "Source data changed since this overview was generated — regenerate the brief to refresh it.", "fr": "Les données sources ont changé depuis la génération de cette vue d'ensemble — régénérez le brief pour l'actualiser.", "es": "Los datos de origen han cambiado desde que se generó este panorama; vuelva a generar el informe para actualizarlo.", "de": "Die Quelldaten haben sich seit der Erstellung dieses Überblicks geändert — erstellen Sie den Brief erneut, um ihn zu aktualisieren."},
+    "overview_empty_title": {"en": "No company overview yet", "fr": "Aucune vue d'ensemble pour le moment", "es": "Aún no hay panorama de la empresa", "de": "Noch kein Unternehmensüberblick"},
+    "overview_empty_body": {"en": "Generate a brief from the sidebar — the overview is produced in the same run.", "fr": "Générez un brief depuis la barre latérale — la vue d'ensemble est produite au cours de la même exécution.", "es": "Genere un informe desde la barra lateral; el panorama se produce en la misma ejecución.", "de": "Erstellen Sie über die Seitenleiste einen Brief — der Überblick wird im selben Durchlauf erzeugt."},
     "primer_loading_market": {"en": "Loading current market context…", "fr": "Chargement du contexte de marché actuel…", "es": "Cargando el contexto actual de mercado…", "de": "Aktueller Marktkontext wird geladen…"},
-    "primer_backfill_warning": {"en": "Company-profile source coverage is incomplete. Run `python ingest.py {ticker}` to backfill the 10-K Business and segment/geography sections.", "fr": "La couverture des sources du profil est incomplète. Lancez `python ingest.py {ticker}` pour récupérer les sections Business et segments/géographies du 10-K.", "es": "La cobertura de fuentes del perfil está incompleta. Ejecute `python ingest.py {ticker}` para recuperar las secciones Business y segmentos/geografía del 10-K.", "de": "Die Quellenabdeckung des Unternehmensprofils ist unvollständig. Führen Sie `python ingest.py {ticker}` aus, um die 10-K-Abschnitte Business und Segmente/Geografie nachzuladen."},
+    "primer_backfill_warning": {"en": "Company Overview source coverage is incomplete. Run `python ingest.py {ticker}` to backfill the 10-K Business and segment/geography sections.", "fr": "La couverture des sources de la vue d'ensemble est incomplète. Lancez `python ingest.py {ticker}` pour récupérer les sections Business et segments/géographies du 10-K.", "es": "La cobertura de fuentes del panorama de la empresa está incompleta. Ejecute `python ingest.py {ticker}` para recuperar las secciones Business y segmentos/geografía del 10-K.", "de": "Die Quellenabdeckung des Unternehmensüberblicks ist unvollständig. Führen Sie `python ingest.py {ticker}` aus, um die 10-K-Abschnitte Business und Segmente/Geografie nachzuladen."},
+    "primer_backfill_warning_hosted": {"en": "This deployment's data snapshot does not yet include the 10-K Business and segment/geography sections for {ticker}. The maintainer needs to re-ingest locally and redeploy the data — this cannot be fixed from within the app.", "fr": "Le snapshot de données de ce déploiement ne contient pas encore les sections Business et segments/géographies du 10-K pour {ticker}. Le mainteneur doit relancer l’ingestion en local et redéployer les données — ce problème ne peut pas être corrigé depuis l’application.", "es": "La instantánea de datos de este despliegue aún no incluye las secciones Business y segmentos/geografía del 10-K para {ticker}. El responsable debe volver a ejecutar la ingesta localmente y redesplegar los datos; este problema no se puede corregir desde la aplicación.", "de": "Der Daten-Snapshot dieses Deployments enthält die 10-K-Abschnitte Business und Segmente/Geografie für {ticker} noch nicht. Der Betreiber muss die Daten lokal neu einlesen und erneut bereitstellen; dies kann nicht innerhalb der App behoben werden."},
     "primer_identity_unavailable": {"en": "The business description is not yet available from verified sources.", "fr": "La description de l’activité n’est pas encore disponible dans les sources vérifiées.", "es": "La descripción del negocio aún no está disponible en fuentes verificadas.", "de": "Die Geschäftsbeschreibung ist aus verifizierten Quellen noch nicht verfügbar."},
     "primer_business": {"en": "How it makes money", "fr": "Comment l’entreprise gagne de l’argent", "es": "Cómo gana dinero", "de": "Wie das Unternehmen Geld verdient"},
     "primer_business_unavailable": {"en": "No verified business-line disclosure is available.", "fr": "Aucune ventilation vérifiée des activités n’est disponible.", "es": "No hay desglose verificado de líneas de negocio.", "de": "Keine verifizierte Aufschlüsselung der Geschäftsbereiche verfügbar."},

dashboard/nav.py

@@ -1,7 +1,7 @@
 """Data-driven sidebar navigation for Amplegest — institutional style.
 
-Five primary destinations, from company onboarding to decision and audit:
-  Company Primer → PM Flash → Ask AI → Evidence & Deltas → Financials
+Five primary destinations, from company onboarding to decision, audit, and tooling:
+  Company Overview → PM Flash → Evidence & Deltas → Financials → Ask AI
 
 Each item has a stable ``key`` decoupled from its display label,
 so renaming a section never breaks routing.
@@ -24,14 +24,14 @@ SECTION_COLORS: dict[str, tuple[str, str, str]] = {
 }
 
 # ── Navigation items ──────────────────────────────────────────────────────────
-# Company Primer onboards the analyst; Chat stays immediately after PM Flash
-# for fast interrogation of the decision evidence.
+# Company Overview onboards the analyst; the violet cross-cutting AI tool stays
+# last, after the decision, evidence, and financial-data screens.
 NAV_ITEMS: list[dict] = [
     {"key": "company",    "i18n_key": "nav_company"},
-    {"key": "verdict",    "i18n_key": "nav_verdict", "display_label": "PM Flash"},
-    {"key": "chat",       "i18n_key": "nav_chat"},
-    {"key": "signals",    "i18n_key": "nav_signals", "display_label": "Evidence & Deltas"},
+    {"key": "verdict",    "i18n_key": "nav_verdict"},
+    {"key": "signals",    "i18n_key": "nav_signals"},
     {"key": "financials", "i18n_key": "nav_financials"},
+    {"key": "chat",       "i18n_key": "nav_chat"},
 ]
 
 VALID_KEYS: frozenset[str] = frozenset(item["key"] for item in NAV_ITEMS)

tests/test_company_profile.py

@@ -10,10 +10,12 @@ from agent.company_profile import (
     _collect_translation_segments,
     _prune_unverified,
     _quarantine_unsupported_shares,
+    collect_profile_evidence,
+    finalize_profile_from_synthesis,
     profile_source_coverage,
     source_fingerprint,
 )
-from agent.evidence import make_evidence_record
+from agent.evidence import evidence_envelope, make_evidence_record
 from agent.company_profile_schemas import CompanyProfile
 from analysis.company_attention import attach_attention_stats, cluster_questions
 from analytics.company_market import (
@@ -199,6 +201,62 @@ def test_company_profile_schema_accepts_partial_payload():
     assert profile.business_lines == []
 
 
+def test_finalize_profile_from_synthesis_prunes_unverified(monkeypatch):
+    record = make_evidence_record(
+        source="10-K",
+        content="The company sells verified products.",
+        document_id="sec:AAPL:profile",
+        chunk_id="business:0",
+    )
+    section = {
+        "business_lines": [{
+            "name": "Unsupported line",
+            "description": {
+                "text": "An unsupported business line.",
+                "source": "10-K",
+                "reliability": "HIGH",
+                "evidence_snippet": "This snippet was never retrieved.",
+                "evidence_ref": record.ref.model_dump(mode="json"),
+            },
+        }],
+    }
+    monkeypatch.setattr(
+        "agent.company_profile.metrics_db.get_all_metrics",
+        lambda ticker: [{"company_name": "Apple Inc.", "filing_date": "2026-01-01"}],
+    )
+    monkeypatch.setattr("agent.company_profile.source_fingerprint", lambda ticker: "fingerprint")
+    monkeypatch.setattr(
+        "agent.company_profile.profile_source_coverage",
+        lambda ticker: {"status": "COMPLETE"},
+    )
+    monkeypatch.setattr("agent.company_profile._annual_trends", lambda ticker: [])
+    monkeypatch.setattr(
+        "analysis.company_attention.attach_attention_stats",
+        lambda profile, ticker, news: profile,
+    )
+
+    profile = finalize_profile_from_synthesis("AAPL", section, [], "test-model")
+
+    assert profile["business_lines"] == []
+    assert profile["status"] == "PARTIAL"
+    assert profile["verification_report"]["removed"] >= 1
+
+
+def test_collect_profile_evidence_can_skip_metrics(monkeypatch):
+    called = []
+
+    def fake_invoke(tool, arguments):
+        called.append(tool.name)
+        return evidence_envelope(tool=tool.name, status="EMPTY", query=arguments)
+
+    monkeypatch.setattr("agent.company_profile._invoke_tool", fake_invoke)
+
+    payloads = collect_profile_evidence("AAPL", include_metrics=False)
+
+    assert len(payloads) == 4
+    assert "get_financial_metrics" not in called
+
+
 def test_prune_unverified_removes_nested_profile_claims():
     profile = {
         "identity": {

tests/test_graph.py

@@ -8,7 +8,7 @@ import json
 
 from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
 
-from agent.evidence import evidence_envelope, make_evidence_record
+from agent.evidence import evidence_envelope, make_evidence_record, parse_evidence_envelope
 from agent.graph import (
     should_continue,
     nudge_node,
@@ -20,6 +20,10 @@ from agent.graph import (
     MAX_FILING_SIGNALS,
     MAX_TRANSCRIPT_SIGNALS,
     _coverage_report,
+    _finalize_synthesis_profile,
+    _partial_brief,
+    _pop_company_profile,
+    profile_evidence_node,
 )
 
 
@@ -30,6 +34,7 @@ def _state(messages, tool_round_count, nudge_fired: bool = False) -> AgentState:
         "tool_round_count": tool_round_count,
         "nudge_fired": nudge_fired,
         "edge_signals": None,
+        "profile_payloads": None,
         "language": "English",
         "brief": None,
         "brief_markdown": None,
@@ -106,6 +111,153 @@ def _error_tool_msg(name: str) -> ToolMessage:
     )
 
 
+def test_profile_evidence_node_injects_parseable_envelopes(monkeypatch):
+    payloads = [
+        evidence_envelope(
+            tool="search_filing",
+            records=[make_evidence_record(
+                source="10-K",
+                content=f"Profile evidence {index}.",
+                document_id=f"sec:AAPL:profile:{index}",
+                chunk_id=str(index),
+            )],
+        )
+        for index in range(2)
+    ]
+    monkeypatch.setattr(
+        "agent.company_profile.collect_profile_evidence",
+        lambda ticker, include_metrics=True: payloads,
+    )
+
+    result = profile_evidence_node(_state([], 0))
+
+    assert result["profile_payloads"] == payloads
+    assert len(result["messages"]) == 3
+    assert all(isinstance(message, HumanMessage) for message in result["messages"])
+    assert result["messages"][0].content == (
+        "== COMPANY PROFILE EVIDENCE "
+        "(deterministic retrieval, evidence.v1 envelopes follow) =="
+    )
+    assert parse_evidence_envelope(result["messages"][0]) is None
+    assert all(
+        parse_evidence_envelope(message) is not None
+        for message in result["messages"][1:]
+    )
+    assert [message.content for message in result["messages"][1:]] == payloads
+
+    monkeypatch.setattr(
+        "agent.company_profile.collect_profile_evidence",
+        lambda ticker, include_metrics=True: [],
+    )
+    assert profile_evidence_node(_state([], 0)) == {
+        "profile_payloads": [],
+        "messages": [],
+    }
+
+
+def test_partial_brief_has_company_profile_none():
+    partial = _partial_brief(_state([], 0), "No usable evidence.")
+    assert partial["company_profile"] is None
+
+
+def test_synthesis_pops_company_profile_before_validation(monkeypatch):
+    from agent.post_synthesis import apply_reliability
+    from agent.schemas import BriefOutput
+
+    record = make_evidence_record(
+        source="10-Q",
+        content="Revenue increased due to higher demand.",
+        document_id="sec:AAPL:brief",
+        chunk_id="mda:0",
+        as_of="2026-04-30",
+    )
+    fact = {
+        "text": "Revenue increased due to higher demand.",
+        "source": "10-Q",
+        "reliability": "HIGH",
+        "evidence_snippet": "Revenue increased due to higher demand.",
+        "evidence_ref": record.ref.model_dump(mode="json"),
+    }
+    payload = evidence_envelope(tool="search_filing", records=[record])
+    data = {
+        "ticker": "AAPL",
+        "company_name": "Apple Inc.",
+        "filing_date": "2026-04-30",
+        "what_matters_most": "Verified demand evidence is the central fact.",
+        "standout_number": fact,
+        "what_changed": [],
+        "bull_points": [],
+        "bear_points": [],
+        "what_to_watch": [],
+        "trends": [],
+        "mda_summary": {
+            "drivers": [],
+            "headwinds": [],
+            "language_shift": "No verified cross-period shift.",
+            "key_quote": fact,
+        },
+        "risks_categorized": [],
+        "management_commentary": [],
+        "guidance_history": [],
+        "company_profile": {
+            "business_lines": [{
+                "name": "Unsupported",
+                "description": {
+                    **fact,
+                    "text": "This profile fact is not in the record.",
+                },
+            }],
+        },
+    }
+
+    profile_section = _pop_company_profile(data)
+    brief = BriefOutput.model_validate(data)
+    verified = apply_reliability(brief.model_dump(), evidence_payloads=[payload])
+
+    assert "company_profile" not in data
+    assert profile_section["business_lines"][0]["name"] == "Unsupported"
+    assert verified["evidence_coverage"] == {
+        "status": "VERIFIED",
+        "verified": 2,
+        "unverified": 0,
+        "failed": 0,
+        "total": 2,
+    }
+
+    finalized = {"ticker": "AAPL", "status": "PARTIAL"}
+    calls = {}
+
+    def fake_finalize(ticker, section, payloads, model):
+        calls["finalize"] = (ticker, section, payloads, model)
+        return finalized
+
+    def fake_save(ticker, profile):
+        calls["save"] = (ticker, profile)
+
+    monkeypatch.setattr(
+        "agent.company_profile.finalize_profile_from_synthesis", fake_finalize
+    )
+    monkeypatch.setattr("storage.company_profiles.save_profile", fake_save)
+    state = _state([_tool_msg("search_filing")], 1)
+    state["profile_payloads"] = [payload]
+
+    assert _finalize_synthesis_profile(state, profile_section, "test-model") == finalized
+    assert calls["finalize"][0] == "AAPL"
+    assert calls["finalize"][2] == state["messages"] + [payload]
+    assert calls["save"] == ("AAPL", finalized)
+
+    monkeypatch.setattr(
+        "agent.company_profile.finalize_profile_from_synthesis",
+        lambda *args: (_ for _ in ()).throw(RuntimeError("profile failed")),
+    )
+    before_failure = {k: v for k, v in verified.items() if k != "company_profile"}
+    verified["company_profile"] = _finalize_synthesis_profile(
+        state, profile_section, "test-model"
+    )
+    assert verified["company_profile"] is None
+    assert {key: value for key, value in verified.items() if key != "company_profile"} == before_failure
+
+
 # ── Cap-based routing (existing tests, renamed) ──────────────────────────────
 
 

tests/test_prompts.py

@@ -1,4 +1,4 @@
-from agent.prompts import SYNTHESIS_STRUCTURED_PROMPT
+from agent.prompts import PROSE_FIELDS, SYNTHESIS_STRUCTURED_PROMPT
 
 
 def _json_template_block() -> str:
@@ -32,3 +32,20 @@ def test_guidance_history_example_does_not_contradict_the_null_out_rule():
     guidance_block = block[guidance_start:block.index("],", guidance_start) + 2]
     assert '"actual_result": null' in guidance_block
     assert '"verdict": null' in guidance_block
+
+
+def test_company_profile_contract_is_in_synthesis_prompt():
+    assert "## COMPANY PROFILE SECTION" in SYNTHESIS_STRUCTURED_PROMPT
+    assert '"company_profile"' in _json_template_block()
+    for field in (
+        "identity", "business_lines", "geographic_exposures",
+        "strategic_changes", "attention_themes", "watch_variables",
+    ):
+        assert f'"{field}"' in _json_template_block()
+
+
+def test_company_profile_analytical_prose_fields_are_translatable():
+    assert {
+        "economics", "why_it_matters", "theme", "variable",
+        "next_datapoint", "alert_signal",
+    } <= PROSE_FIELDS

tests/test_verdict.py

@@ -8,6 +8,7 @@ from dashboard.verdict import (
     _select_swing_factor,
 )
 from dashboard.signals_view import _sentiment_display_allowed
+from dashboard.i18n import STRINGS, t
 from dashboard.nav import NAV_ITEMS, VALID_KEYS
 
 
@@ -125,12 +126,20 @@ def test_sentiment_display_policy_fails_closed():
     assert _sentiment_display_allowed({"display_policy": {"sentiment_calibrated": True}}) is True
 
 
-def test_pm_navigation_preserves_route_keys_and_promotes_chat():
-    assert VALID_KEYS == {"verdict", "signals", "financials", "chat"}
-    chat = next(item for item in NAV_ITEMS if item["key"] == "chat")
-    # Chat is promoted to primary nav, right after the decision view (PM Flash) —
-    # no longer a "secondary" item relegated under an "Evidence tool" heading.
-    assert "secondary" not in chat
-    assert NAV_ITEMS[0]["key"] == "verdict"
-    assert NAV_ITEMS[1]["key"] == "chat"
+def test_nav_order_is_analyst_flow():
+    assert VALID_KEYS == {"company", "verdict", "signals", "financials", "chat"}
+    assert [item["key"] for item in NAV_ITEMS] == [
+        "company", "verdict", "signals", "financials", "chat"
+    ]
     assert not any(item.get("secondary") for item in NAV_ITEMS)
+
+
+def test_nav_items_have_no_hardcoded_display_label():
+    assert not any("display_label" in item for item in NAV_ITEMS)
+
+
+def test_nav_i18n_labels_complete():
+    for item in NAV_ITEMS:
+        labels = STRINGS[item["i18n_key"]]
+        assert set(labels) == {"en", "fr", "es", "de"}
+    assert t("nav_company") == "Company Overview"