fix: replace exact-substring evidence_snippet match with fuzzy word-run match

fix: per-item drop of invalid brief facts; relax LLM-produced singleton fields

fix: tolerate malformed LLM-copied EvidenceRef fields

fix: normalize off-list enum values in company profile schemas

fix: allow metrics/analyst sources in CategorizedRisk and GuidancePoint

fix: raise synthesis max_tokens to 64000 and surface truncation explicitly

fix: expose CHECK_REQUIRED periods as field-level-disclosed evidence instead of hiding entire rows

fix: keep fail-closed per fact but stop zero-verified synthesis from emptying the whole brief

feat: merge Company Overview generation into single synthesis call

feat: per-fact evidence_ref enforcement + AAPL delta re-ingestion

feat: add company primer research tab

feat: multi-provider LLM support, prominent chat, design pass, and new analytics

fix: move empty-path guard before loop in _set_at_path

feat: add agent/translate.py — segment-map brief translation

feat: embed synthesis language in brief dict

feat: extract shared prose-field constants in agent/prompts.py

feat: add since_release_price_reaction_pct — cumulative return from earnings release to today

chore: checkpoint sidebar nav + i18n work-in-progress before Decision Stack redesign

fix: coerce tax_rate_and_effective_leverage to consensus_beat_mix + harden prompts

fix: correct tells count in READING BETWEEN THE LINES header (four → five)

fix: accounting_quality tell + anchoring + routing clarification in prompts

fix: normalize signal_type case-insensitively + fix subtext typo

feat: add READING BETWEEN THE LINES synthesis pass to prompts

feat: add SubtextRead + between_the_lines to BriefOutput schema

fix: harden _extract_json — use raw_decode to ignore trailing JSON objects and prose

feat: add Chat tab — bounded Q&A agent over stored filings and transcripts

feat: add average_brief_cost_usd helper to cost_log for portfolio cost estimation

fix: coerce compound LLM source strings to prevent BriefOutput validation failures

feat: Analyst Edge layer — deterministic text delta signals (Phase 0+1)

fix: insert stub ToolMessages when round cap hit mid-tool-call

fix: verdict layout, N/A market pulse, impact badge colors

feat: add impact badge to all sourced facts

fix: retry on overloaded_error + friendlier API error messages

feat: analytical edge section + fix HF Spaces fragment polling race

fix: background generation, parallel tools, prompt caching, Tavily cache

deploy: Primer app for HF Spaces — clean orphan history