"בית ספר להחלטות" Phase 2 — the system now has formal analytical
methodology for building quasi-judicial decisions, separate from
Dafna's writing style (SKILL.md) and content checklists.
What was done:
- Downloaded 5 authoritative sources (~341K words): FJC Judicial
Writing Manual (1991+2020), Garner Legal Writing in Plain English,
Posner How Judges Think, Scalia/Garner Making Your Case
- Extracted principles from all sources into intermediate docs
- Synthesized into docs/decision-methodology.md (3,400 words,
12 sections, 10 guiding principles)
- Integrated methodology into block-yod prompt via {methodology_guidance}
- Restructured legal-writer agent workflow to follow analytical stages
- Made "answer all claims" flexible (bundle/skip via chair_directions)
- Added methodology compliance check (#7) to legal-qa agent
- Updated all knowledge files (CLAUDE.md, SKILL.md, lessons, corpus)
Three-layer architecture:
1. Methodology (decision-methodology.md) — universal, how to think
2. Content checklists (lessons.py) — specific per appeal subtype
3. Style (SKILL.md) — Dafna's personal writing patterns
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Addresses Dafna's observation that licensing decisions lack comprehensive
planning discussion. Systematic corpus analysis of all 24 training decisions
revealed the system learned writing style but not substantive content.
Changes:
- Corpus analysis of all 24 decisions (docs/corpus-analysis.md)
- 5 content checklists by appeal subtype injected into block-yod prompt
- chair_feedback DB table + API endpoints + MCP tools
- Feedback management page in Next.js UI (/feedback)
- Navigation updated with "הערות יו״ר" link
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Closes the loop so דפנה's positions (written inline in the UI and
saved to analysis-and-research.md) automatically become binding
direction for the legal-writer agent — no manual copy-paste,
no bypass.
Backend:
- research_md.extract_chair_directions(path) returns a compact dict
with status (missing/empty/partial/complete), filled_count,
empty_count, and a reduced list of threshold_claims + issues each
with {id, number, title, direction}. Designed to be directly usable
as direction_doc by the writer.
- New MCP tool: drafting.get_chair_directions(case_number) wraps the
helper, resolves the case research file path via config.find_case_dir,
returns formatted JSON.
- Registered in server.py as mcp__legal-ai__get_chair_directions.
legal-writer agent update:
- Adds get_chair_directions to the tools list.
- New mandatory "שלב 1ב" before any block writing: call
get_chair_directions, branch on status.
- missing → halt, report "legal-analyst לא רץ עדיין"
- empty → halt, instruct Dafna to fill positions via the UI URL
- partial → halt unless user confirms; write only filled sections
- complete → proceed
- New "שלב 1ג" constructs an internal direction_doc from the
received chair rulings before writing block י.
- Block י section expanded with 5 binding rules:
1. Open each discussion with Dafna's ruling as the thesis
2. Frame the reasoning in her style (use get_style_guide phrases)
3. Match her tone (decisive vs nuanced)
4. Must NOT contradict her position — if she disagreed with your
own inclination, her position rules
5. Use legal_questions from the analysis file as the analytical
structure (principle question first, concrete application second)
- New bullet section for block יא: summarize each chair ruling
briefly, state final outcome, close with the signed date formula.
Verified all four status paths (missing/empty/partial/complete) via
local test. Now Dafna's workflow is fully end-to-end: she reads the
analyst report in the UI, fills "עמדת ועדת הערר" in each card, hits
blur to auto-save, then triggers legal-writer — which picks up her
positions as direction without any file shuffle.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New feature on case view: the analysis-and-research.md produced by the
legal-analyst agent is now rendered as structured cards in the UI,
with inline editing of "עמדת ועדת הערר" that writes directly back to
the markdown file (atomic rename).
Backend (research_md.py):
- parse(Path) → dict with header, prose sections, threshold_claims[],
issues[], conclusions, other_sections
- Tolerant field extractor handles both block ("**LABEL:**\ncontent")
and inline ("**LABEL:** content") variants
- Detects [ימולא ע"י יו"ר הוועדה] placeholder → empty chair_position
- update_chair_position(path, section_id, text) locates the exact
subsection by ordinal, replaces or appends the chair field, writes
atomically via temp file + os.replace
- Section IDs: threshold_N / issue_N (1-based)
Endpoints:
- GET /api/cases/{n}/research/analysis — returns parsed JSON or 404
- PATCH /api/cases/{n}/research/analysis/chair-position — {section_id, position}
Frontend (#page-case):
- New card "ניתוח משפטי ומחקר" below local-files card
- Prose sections as justified text panels (background + gold border)
- Threshold claims and issues as collapsible <details> items with
gold right-border on open, numbered pills
- Each item shows all extracted fields with label above content
- Chair position editor: gold-wash background, 📝 icon label, textarea
with placeholder prompt
- onblur → PATCH with save indicator: ⏳ שומר → ✓ נשמר HH:MM → fade
- Status pill next to each item title: "ממתין לעמדה" / "✓ עמדה נקבעה"
- First threshold claim opens by default, rest closed
- Card hidden entirely when no analysis file exists (404)
Tested against real file: case 1033-25 with 3 threshold claims and
6 issues, all chair positions correctly empty, update writes only the
targeted section, atomic rewrite preserves all other content.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- DELETE /api/training/corpus/{id} + delete button on training page,
with confirmation dialog and recompute hint
- /api/system/tasks + floating process panel (bottom-left) showing
active background tasks with live 3s polling
- /api/system/recent-activity derives a feed from cases, style_corpus,
and last style_patterns run; sidebar on home page renders with
relative timestamps
- /api/system/diagnostics + /#/diagnostics page showing DB health,
row counts per table, active tasks, stuck documents (>10 min),
failed extractions
- Cosmetic: signature phrase headline now prefers clean phrases over
bracket-heavy templates for display
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- New proofreader service strips Nevo editorial additions (front matter,
postamble, page headers, watermarks, inline codes) from DOCX/PDF/MD
- PDF pages use Google Vision OCR for clean Hebrew RTL extraction
- New training page at #/training with drag-and-drop upload, automatic
metadata extraction (decision number, date, categories), reviewable
preview, and style pattern report grouped by type
- API endpoints: /api/training/{analyze,upload,corpus,patterns,
analyze-style,analyze-style/status}
- Fix claude_session.query to pipe prompt via stdin, avoiding ARG_MAX
overflow when analyzing 900K+ char corpus
- CLI scripts for batch proofreading and corpus upload
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add delete_document_chunks for reprocessing, save extracted text to disk
- Expand case directory structure (original/extracted/proofread/backup)
- Update classifier patterns (תגובה, הודעת עמדה)
- Fix proofreader agent paths for new directory layout
- Update HEARTBEAT to notify on every task completion
- Improve bidi_table with LRE/PDF directional embedding
- Add Paperclip project verification and auto-close setup issue
- Add auto-sync-cases.sh for Gitea synchronization
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Remove cases/new|in-progress|completed subdivision (status managed in DB)
- Rename documents/original → documents/originals (consistent plural)
- Move exports from global data/exports/ into cases/{num}/exports/
- Add documents/research/ for case law and analysis files
- Update all agents, scripts, config, web API endpoints, and DB paths
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Benchmark on case 1130-25 (4 Hebrew legal docs, 8 queries) showed:
- voyage-law-2: avg top-1 score 0.5839 (+27% over voyage-3-large)
- voyage-4-large: avg top-1 score 0.4119 (worse than current)
- voyage-3-large: avg top-1 score 0.4589 (baseline)
voyage-law-2 costs ~4.6x more per run but delivers significantly
better retrieval quality for Hebrew legal text. Model is now
configurable via VOYAGE_MODEL env var.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Export DOCX now saves to data/exports/{case_number}/ with auto-versioning
(טיוטה-v1, v2...). The case view UI shows all drafts with download buttons,
allows uploading revised versions (עריכה-v1...), and marking a version as
final (copies to training corpus for style learning).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two issues that caused QA agent to fail:
1. save_block_content saved to DB only — now also rebuilds drafts/decision.md
2. legal-writer.md now has explicit mandatory step: case_update(status="drafted")
Without these, workflow_status reports has_draft=false and QA can't run.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New module claude_session.py provides query() and query_json() that
run prompts via `claude -p` CLI — uses the claude.ai session, zero API cost.
Converted 6 services:
- claims_extractor.py: extract_claims_with_ai
- brainstorm.py: brainstorm_directions
- block_writer.py: write_block (was streaming+thinking, now simple)
- qa_validator.py: claims_coverage check
- style_analyzer.py: 3 API calls (single pass, multi pass, synthesis)
- learning_loop.py: extract_lessons
Only extractor.py still uses Anthropic API (for PDF OCR with Vision).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replaces API-based classifier with:
1. Filename pattern matching (covers 95%+ of legal docs)
2. Content keyword matching for ambiguous filenames
3. Claude Code headless (claude -p) fallback for edge cases
No Anthropic API calls needed for classification.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Text extraction, chunking and embedding proceed even if Claude API
classification or reference extraction fails (e.g. API quota exceeded).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace single CASES_DIR with find_case_dir() that searches across
all status directories. New cases created in cases/new/{number}/.
Config: CASES_BASE, CASES_NEW, CASES_IN_PROGRESS, CASES_COMPLETED
Docker: added -v /home/chaim/legal-ai/cases:/cases volume mount
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
classify_document and identify_parties both used Haiku, which produced
parsing failures and 0% confidence on Beit HaKerem documents.
Sonnet handles Hebrew legal documents more reliably.
No more Haiku usage in the entire codebase.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two fixes for claims_coverage false negatives (55% → expected ~85%+):
1. Model upgrade: Haiku → Sonnet for semantic matching. Haiku missed
obvious matches (e.g., paragraph about "כריתת עצים" not matching
claim about tree cutting). Sonnet understands context better.
2. Filter: only check appellant/respondent claims, not committee or
permit_applicant claims. Committee claims are defensive positions
("the application complies with the plan") — they don't need to
be "addressed" in the discussion section.
3. Send full discussion text (was truncated to 12K chars).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New architecture: MCP provides context, Claude Code writes.
New functions:
- get_block_context(case_id, block_id) → returns full context package
(prompt, source docs, claims, direction, precedents, style guide)
WITHOUT calling Anthropic API
- save_block_content(case_id, block_id, content) → saves block to DB
New MCP tools: get_block_context, save_block_content
The old write_block (API-based) still works as fallback.
The new flow uses Claude Code's own model (Opus 4.6, 1M context)
which has no separate API billing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
_build_style_context rewritten from 10-line summary to comprehensive
style guide including:
- Tone rules per appeal type (warm for licensing, cold for levy)
- 15 mandatory expressions ("כידוע", "ברי כי", "אין בידנו לקבל")
- Discussion structure rules (continuous prose, conclusion first)
- Per-party phrasing templates (appellants, committee, permit applicants)
- DB patterns grouped by type (phrases, transitions, openings, closings)
This addresses the main quality gap: style rated 2/5 because the output
was "dry and overly formal" vs Dafna's "direct and clear" voice.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
config.py parse_llm_json: Added truncated JSON recovery. When Claude's
output is cut mid-JSON (common with long claim lists), the parser now:
- Finds the last complete JSON item (closing "}")
- Closes the array/object brackets
- Returns partial but valid results instead of None
Tested: recovers 2/3 items from truncated array, all cases pass.
claims_extractor.py:
- Prompt asks for compact output (150 words max per claim, group similar)
- Explicitly requests "no markdown, no explanations, JSON only"
- Long documents split into chunks at paragraph boundaries
- Each chunk processed separately, results merged
- max_tokens already at 8192
This fixes the recurring "0 claims" bug for committee responses and
permit applicant responses where the JSON was getting truncated.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
block-yod max_tokens reduced from 32K to 16K — the API returned
"max_tokens: 32768 > 32000" error. With thinking enabled, the actual
limit for output is lower. 16K is sufficient for discussion blocks.
Also: extractor.py now supports .md files (was missing, blocked
Beit HaKerem upload).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
block_writer: Rewrote block-zayin prompt to require synthesis by topic
instead of listing each claim separately. Now produces 3 organized
sections (appellants 8, committee 6, permit applicants 3+) instead
of 40 scattered paragraphs. Target: 800-1500 words.
claims_extractor: Fix markdown code block stripping (same bug as
qa_validator had). Enables parsing claims from Claude responses
wrapped in ```json blocks.
Tested on Hecht: block-zayin from 40 paragraphs/1049 words to
17 organized paragraphs/1039 words. Structure now matches Dafna's
original (3 parties, grouped by topic).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
block_writer: new renumber_all_blocks() function that renumbers all
paragraphs across all blocks sequentially (1, 2, 3...). Handles both
plain "N." and bold "**N.**" formats. Added missing 'import re'.
qa_validator: sequential_numbering check now matches bold-formatted
numbers (**N.**) in addition to plain (N.).
Tested on Hecht: renumbered 115 paragraphs across 7 blocks, QA 6/6.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
block_writer: _build_precedents_context now searches both
paragraph_embeddings (other decisions by Dafna) and case_law_embeddings
(precedent case law). Previously only searched document_chunks which
had no cross-case data. Now returns ~2400 chars from 3 other decisions.
processor: Step 1.6 auto-updates case appellants/respondents from
classifier results when they're empty. Prevents blank party fields.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
claims_coverage now uses Claude Haiku to check if each claim is
semantically addressed in the discussion, not just keyword-matched.
- Sends all claims + discussion to Claude in one API call
- Returns addressed/partial/missing for each claim
- Handles markdown code block wrapping in response
- max_tokens 4096 (was 2048) for 48+ claims
Result on Hecht: 45/48 addressed (94%), 1 partial, 3 missing.
The 3 missing are genuinely unaddressed (personal/procedural claims).
Previously keyword check showed 47/48 but missed semantic gaps.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
block_writer: _build_claims_context now filters out block-zayin claims
(from final decision) and uses only claims from original pleadings.
Reduces noise from 78 to 48 real claims for Hecht case.
qa_validator: claims_coverage check rewritten:
- Filter block-zayin claims (same reason)
- Keyword-based matching instead of 3-word phrase matching
- 25% keyword overlap threshold (was: any 3-word match)
- Allow up to 20% uncovered claims before failing
- Check both block-yod and block-zayin for coverage
Result: Hecht case QA goes from 4/6 to 6/6, 47/48 claims covered (98%).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add minimum word count guidance (2000-4000 words)
- Number each claim in claims_context for explicit tracking
- Require 3-5 case law citations minimum
- Fix max_tokens > budget_tokens for extended thinking
- Use streaming for opus+thinking requests (>10min timeout)
Tested on Hecht case: block-yod improved from 1039 to 1927 words.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Previously matched any word after תב"ע (e.g., "תב"ע ואין", "תב"ע קיפחה").
Now requires a plan number (digits/hyphens) — reduces false positives from 24 to 4
on the Hecht case test.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add expected_outcome field to cases (rejection/partial/full/betterment_levy)
- New lessons.py module with golden ratios, templates, and drafting guidance per outcome type
- Style analyzer now uses Opus with full decision text (no truncation), with multi-pass fallback for large corpora
- Drafting tool provides outcome-specific templates, section guidance, and ratio comments
- Improved JSON extraction with bracket-matching fallback
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Ezer Mishpati - AI legal decision drafting system with:
- MCP server (FastMCP) with document processing pipeline
- Web upload interface (FastAPI) for file upload and classification
- pgvector-based semantic search
- Hebrew legal document chunking and embedding
