fix: Cache-hit whitelist and LangGraph-dropped gate state

fix: Margin-consistency gate check and final-evaluation label

feat: Surface data-gate findings in result quality notes

feat: Deterministic data gate on the Researcher->Analyzer edge

fix: Gateway resilience to Researcher restarts at submit time

fix: Boeing-run findings — canonical refs, edge contracts, guardrails

fix: Resubmit research task once when the service loses it mid-run

feat: JSON-mode structured outputs for critic scoring

feat: Vite 7 + React 19 upgrade and SSE status streaming

fix: Handle gpt-oss reasoning-token budget in Groq calls and deep health

feat: Harden API, pin dependencies, and enable pytest CI gate

chore: Remove AWS deployment, dead code, and unused dependencies

fix: Migrate to living LLM models and defuse stale production config

debug: Add detailed logging for numeric mismatches in critic

fix: Align analyzer field names with MCP data structure

fix: Improve SWOT output formatting

fix: Add MCP 'executing' state for process flow visualization

fix: Handle temporal dict format for revenue_cagr_3yr in analyzer

fix: Clean company name suffixes in stock listings

fix: Update analyzer to handle new MCP source-keyed structure

fix: Handle 'sources' wrapper in fundamentals data structure

fix: Only validate SWOT section for uncited numbers

fix: Update citation patterns to match new SWOT output format

feat: SWOT output as structured text parsed into HTML tables

Fix revision loop exit: change > 3 to >= 3, handle ESCALATE

Fix revision loop: add critique_details to AgentState

Reformat SWOT output: concise 3-column table format

Remove duplicate Company Details from markdown output

UI improvements: metric refs, company details, remove status labels

Decrease quality threshold from 7 to 6

Change OpenRouter model to Gemma 3n E4B IT

Fix OpenRouter model: use deepseek-chat-v3 (was invalid model)

Rotate between LLM providers instead of consecutive retries

Increase LLM retry delays and add fallback delay for all errors

Update default models: Gemini stable + OpenRouter GPT fallback

Add 10s delay between provider fallbacks after rate limit (429)

Fix: Metric extraction to match MCP data structure

Increase exponential backoff from 3s to 5s base

Increase LLM rate limit delays from 5s to 10s

Fix: Handle temporal dict structure in _format_metrics_for_prompt

Fix: Temporal data extraction and frontend date display

Add SWOT output enhancements

Add VADER sentiment analysis for news and reddit

Layer 4: Add deterministic numeric validation in Critic

Add compact activity log for Analyzer reference values

Layer 1: Add metric reference table to prevent LLM hallucination

Fix: Increase rate limit delays and fix revision detection

Fix: Add rate limit handling with retry and delays

Fix: Critic LLM failure now triggers graceful exit

Fix: Track actual data source (SEC vs Yahoo) for fundamentals metrics