feat: fix session_pair_harvester to use role/content format (#91)

- Harvester used old message fields (from/value) but Hermes sessions use role/content
- Import session_reader to normalize conversations properly
- Update extract function to operate on normalized role/content messages
- Change predecessor lookup from "human"/"gpt" to "user"/"assistant"
- Add comprehensive smoke tests (8 tests, all pass)
- Verify extraction from test_sessions: 11 pairs, avg ratio 8.13

scripts/progress_tracker.py

@@ -1,477 +0,0 @@
-#!/usr/bin/env python3
-"""
-Progress Tracker — Pipeline 10.8
-Track improvement metrics over time. Are we getting better?
-
-Metrics tracked:
- 1. Test coverage  — % of Python functions with associated tests (test:source file ratio + line coverage if available)
- 2. Doc coverage   — % of Python callables with docstrings (AST-based)
- 3. Issue close rate — closed / (opened + closed) per week (Gitea API)
- 4. Dep freshness  — % of requirements pinned vs outdated (pip list --outdated)
-
-Output:
- - metrics/snapshots/YYYY-MM-DD.json  — one snapshot per run
- - metrics/TRENDS.md                  — cumulative markdown table
- - stdout summary
-
-Usage:
-    python3 scripts/progress_tracker.py
-    python3 scripts/progress_tracker.py --json
-    python3 scripts/progress_tracker.py --output metrics/TRENDS.md
-
-Weekly cron:
-    0 9 * * 1 cd /path/to/compounding-intelligence && python3 scripts/progress_tracker.py
-"""
-
-import argparse
-import json
-import os
-import re
-import subprocess
-import sys
-from collections import defaultdict
-from datetime import datetime, timezone, timedelta
-from pathlib import Path
-from typing import Any, Dict, List, Optional, Tuple
-
-# ── Configuration ──────────────────────────────────────────────────────────
-
-SCRIPT_DIR = Path(__file__).resolve().parent
-REPO_ROOT = SCRIPT_DIR.parent
-METRICS_DIR = REPO_ROOT / "metrics"
-SNAPSHOTS_DIR = METRICS_DIR / "snapshots"
-TOKEN_PATH = Path.home() / ".config" / "gitea" / "token"
-GITEA_API_BASE = "https://forge.alexanderwhitestone.com/api/v1"
-ORG = "Timmy_Foundation"
-
-# Ensure paths exist
-SNAPSHOTS_DIR.mkdir(parents=True, exist_ok=True)
-
-
-# ── Helpers ─────────────────────────────────────────────────────────────────
-
-def run_cmd(cmd: List[str], cwd: Path = REPO_ROOT) -> str:
-    """Run a shell command and return stdout (stderr merged)."""
-    result = subprocess.run(
-        cmd, capture_output=True, text=True, cwd=cwd, timeout=30
-    )
-    if result.returncode != 0:
-        return ""
-    return result.stdout.strip()
-
-
-def slugify_date(dt: datetime) -> str:
-    return dt.strftime("%Y-%m-%d")
-
-
-def snapshot_path(dt: datetime) -> Path:
-    return SNAPSHOTS_DIR / f"{slugify_date(dt)}.json"
-
-
-def load_snapshots() -> List[Dict[str, Any]]:
-    """Load all existing snapshots sorted by date."""
-    snapshots = []
-    for f in sorted(SNAPSHOTS_DIR.glob("*.json")):
-        try:
-            with open(f) as fp:
-                snapshots.append(json.load(fp))
-        except Exception:
-            continue
-    return snapshots
-
-
-# ── Metric 1: Test Coverage ─────────────────────────────────────────────────
-
-def collect_test_coverage() -> Dict[str, Any]:
-    """
-    Compute test coverage metrics.
-    Counts test_*.py and *_test.py files vs non-test .py source files.
-    Also attempts to read .coverage if present.
-    """
-    all_py = list(REPO_ROOT.rglob("*.py"))
-    
-    source_files = []
-    test_files = []
-    
-    for p in all_py:
-        try:
-            rel_parts = p.relative_to(REPO_ROOT).parts
-        except ValueError:
-            continue
-        
-        # Skip hidden/cache/temp dirs (check only relative parts)
-        if any(part.startswith('.') or part.startswith('__') for part in rel_parts):
-            continue
-        if any(part in ('node_modules', 'venv', '.venv', 'env', '.pytest_cache') for part in rel_parts):
-            continue
-        
-        if p.name.startswith("test_") or p.name.endswith("_test.py"):
-            test_files.append(p)
-        else:
-            source_files.append(p)
-    
-    # Try to get line coverage from .coverage
-    coverage_percent = None
-    coverage_tool = None
-    coverage_file = REPO_ROOT / ".coverage"
-    if coverage_file.exists():
-        try:
-            import coverage  # type: ignore
-            # Use coverage API if available
-            cov = coverage.Coverage(data_file=str(coverage_file))
-            cov.load()
-            total = cov.report()
-            coverage_percent = total if isinstance(total, float) else None
-            coverage_tool = "coverage"
-        except Exception:
-            # Fallback: parse `coverage report` output
-            out = run_cmd(["coverage", "report", "--skip-empty"])
-            if out:
-                for line in out.splitlines():
-                    if "TOTAL" in line:
-                        parts = line.split()
-                        if len(parts) >= 2:
-                            try:
-                                coverage_percent = float(parts[-1].rstrip('%'))
-                                coverage_tool = "coverage"
-                                break
-                            except ValueError:
-                                pass
-    
-    return {
-        "test_files": len(test_files),
-        "source_files": len(source_files),
-        "test_to_source_ratio": round(len(test_files) / len(source_files), 4) if source_files else 0.0,
-        "coverage_tool": coverage_tool,
-        "coverage_percent": coverage_percent,
-    }
-
-
-# ── Metric 2: Doc Coverage ──────────────────────────────────────────────────
-
-def collect_doc_coverage() -> Dict[str, Any]:
-    """
-    Check AST of Python files for docstrings.
-    Returns: callables_total, callables_with_doc, doc_coverage_percent
-    """
-    import ast
-    
-    all_py = list(REPO_ROOT.rglob("*.py"))
-    
-    source_files = []
-    test_files = []
-    
-    for p in all_py:
-        try:
-            rel_parts = p.relative_to(REPO_ROOT).parts
-        except ValueError:
-            continue
-        
-        if any(part.startswith('.') or part.startswith('__') for part in rel_parts):
-            continue
-        if any(part in ('node_modules', 'venv', '.venv', 'env', '.pytest_cache') for part in rel_parts):
-            continue
-        
-        if p.name.startswith("test_") or p.name.endswith("_test.py"):
-            test_files.append(p)
-        else:
-            source_files.append(p)
-    
-    total_callables = 0
-    with_doc = 0
-    
-    for p in source_files + test_files:
-        try:
-            with open(p) as f:
-                tree = ast.parse(f.read(), filename=str(p))
-            for node in ast.walk(tree):
-                if isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):
-                    total_callables += 1
-                    doc = ast.get_docstring(node)
-                    if doc and doc.strip():
-                        with_doc += 1
-        except Exception:
-            continue
-    
-    return {
-        "callables_total": total_callables,
-        "callables_with_doc": with_doc,
-        "doc_coverage_percent": round((with_doc / total_callables * 100) if total_callables else 0.0, 2),
-    }
-
-
-# ── Metric 3: Issue Close Rate ──────────────────────────────────────────────
-
-def collect_issue_metrics() -> Dict[str, Any]:
-    """
-    Use Gitea API to get issue open/close stats for the last 7 days.
-    Returns counts and close rate.
-    """
-    token = ""
-    if TOKEN_PATH.exists():
-        token = TOKEN_PATH.read_text().strip()
-    
-    if not token:
-        return {
-            "opened_last_7d": None,
-            "closed_last_7d": None,
-            "close_rate": None,
-            "total_open": None,
-            "note": "Gitea token not available"
-        }
-    
-    try:
-        from urllib.request import Request, urlopen
-        from urllib.error import HTTPError, URLError
-    except ImportError:
-        return {"error": "urllib not available"}
-    
-    now = datetime.now(timezone.utc)
-    week_ago = now - timedelta(days=7)
-    since = week_ago.strftime("%Y-%m-%d")
-    
-    headers = {"Authorization": f"token {token}"}
-    base_url = f"{GITEA_API_BASE}/repos/{ORG}/compounding-intelligence/issues"
-    
-    try:
-        # Get issues from last 7 days
-        url = f"{base_url}?state=all&since={since}&per_page=100"
-        req = Request(url, headers=headers)
-        with urlopen(req, timeout=15) as resp:
-            issues = json.loads(resp.read())
-        
-        opened = 0
-        closed = 0
-        for issue in issues:
-            created = datetime.fromisoformat(issue["created_at"].replace("Z", "+00:00"))
-            if created >= week_ago:
-                opened += 1
-            if issue.get("state") == "closed":
-                closed_at_str = issue.get("closed_at")
-                if closed_at_str:
-                    closed_at = datetime.fromisoformat(closed_at_str.replace("Z", "+00:00"))
-                    if closed_at >= week_ago:
-                        closed += 1
-        
-        # Total open issues
-        req2 = Request(f"{base_url}?state=open&per_page=1", headers=headers)
-        with urlopen(req2, timeout=15) as resp:
-            total_open = int(resp.headers.get("X-Total-Count", "0"))
-        
-        total = opened + closed
-        close_rate = closed / total if total > 0 else 0.0
-        
-        return {
-            "opened_last_7d": opened,
-            "closed_last_7d": closed,
-            "close_rate": round(close_rate, 4),
-            "total_open": total_open,
-        }
-    except Exception as e:
-        return {
-            "opened_last_7d": None,
-            "closed_last_7d": None,
-            "close_rate": None,
-            "total_open": None,
-            "error": str(e)[:100],
-            "note": "Gitea API unavailable"
-        }
-
-
-# ── Metric 4: Dependency Freshness ─────────────────────────────────────────
-
-def collect_dep_freshness() -> Dict[str, Any]:
-    """
-    Check requirements.txt for outdated dependencies using pip list --outdated.
-    Returns freshness percentage and outdated list.
-    """
-    req_file = REPO_ROOT / "requirements.txt"
-    if not req_file.exists():
-        return {
-            "total_deps": 0,
-            "outdated_deps": 0,
-            "freshness_percent": 100.0,
-            "outdated_list": [],
-            "note": "requirements.txt not found"
-        }
-    
-    # Parse requirements (very simple: take name before comparison op)
-    reqs = []
-    with open(req_file) as f:
-        for line in f:
-            line = line.strip()
-            if not line or line.startswith("#"):
-                continue
-            m = re.match(r"^([a-zA-Z0-9_.-]+)", line)
-            if m:
-                reqs.append(m.group(1))
-    
-    if not reqs:
-        return {"total_deps": 0, "outdated_deps": 0, "freshness_percent": 100.0, "outdated_list": []}
-    
-    # Query pip for outdated packages (may fail if pip not available)
-    outdated_names = set()
-    try:
-        out = run_cmd(["pip", "list", "--outdated", "--format=json"])
-        if out:
-            data = json.loads(out)
-            outdated_names = {item["name"].lower() for item in data}
-    except Exception:
-        pass
-    
-    outdated = [p for p in reqs if p.lower() in outdated_names]
-    total = len(reqs)
-    outdated_count = len(outdated)
-    freshness = round(((total - outdated_count) / total * 100) if total else 100.0, 1)
-    
-    return {
-        "total_deps": total,
-        "outdated_deps": outdated_count,
-        "freshness_percent": freshness,
-        "outdated_list": outdated,
-    }
-
-
-# ── Snapshot & Trends ───────────────────────────────────────────────────────
-
-def take_snapshot() -> Dict[str, Any]:
-    """Collect all metrics and return a snapshot dict."""
-    now = datetime.now(timezone.utc)
-    test_cov = collect_test_coverage()
-    doc_cov = collect_doc_coverage()
-    issues = collect_issue_metrics()
-    deps = collect_dep_freshness()
-    
-    return {
-        "timestamp": now.isoformat(),
-        "date": slugify_date(now),
-        "metrics": {
-            "test_coverage": test_cov,
-            "doc_coverage": doc_cov,
-            "issues": issues,
-            "dependencies": deps,
-        }
-    }
-
-
-def save_snapshot(snapshot: Dict[str, Any]) -> Path:
-    path = snapshot_path(datetime.fromisoformat(snapshot["timestamp"]))
-    with open(path, "w") as f:
-        json.dump(snapshot, f, indent=2)
-    return path
-
-
-def generate_trends(snapshots: List[Dict[str, Any]], output_path: Optional[Path] = None) -> str:
-    """Generate markdown trends table; optionally write to file."""
-    if not snapshots:
-        msg = "# Progress Tracker — Trends\n\nNo snapshots yet. Run `progress_tracker.py` to create the first snapshot."
-        if output_path:
-            output_path.parent.mkdir(parents=True, exist_ok=True)
-            output_path.write_text(msg)
-        return msg
-    
-    lines = [
-        "# Progress Tracker — Trends",
-        f"\nLast updated: {datetime.now(timezone.utc).strftime('%Y-%m-%d %H:%M UTC')}",
-        f"\nSnapshots: {len(snapshots)}\n",
-        "| Date | Test Files → Source | Doc Coverage | Issues Closed/Opened (7d) | Dep Freshness |",
-        "|------|---------------------|--------------|---------------------------|---------------|",
-    ]
-    
-    for snap in reversed(snapshots):  # chronological
-        date = snap["date"]
-        m = snap["metrics"]
-        tc = m["test_coverage"]
-        test_str = f"{tc['test_files']}/{tc['source_files']} ({tc['test_to_source_ratio']:.2f})"
-        doc_str = f"{m['doc_coverage']['doc_coverage_percent']:.1f}%"
-        issues_str = f"{m['issues'].get('closed_last_7d','-')}/{m['issues'].get('opened_last_7d','-')}"
-        dep_str = f"{m['dependencies'].get('freshness_percent','?')}%"
-        lines.append(f"| {date} | {test_str} | {doc_str} | {issues_str} | {dep_str} |")
-    
-    # Current snapshot summary
-    cur = snapshots[-1]
-    cm = cur["metrics"]
-    lines.append(f"\n## Current Snapshot ({cur['date']})\n")
-    
-    tc = cm["test_coverage"]
-    cov_line = f"- Test coverage: {tc['coverage_percent']:.1f}% (via {tc['coverage_tool']})\n" if tc["coverage_percent"] else "- Test coverage: (pytest-cov not configured)\n"
-    lines.append(cov_line)
-    lines.append(f"- Doc coverage: {cm['doc_coverage']['doc_coverage_percent']:.1f}%")
-    
-    im = cm["issues"]
-    if im.get("close_rate") is not None:
-        lines.append(f"- Issue close rate (7d): {im['close_rate']*100:.1f}% ({im['closed_last_7d']} closed, {im['opened_last_7d']} opened)")
-    else:
-        lines.append(f"- Issue metrics: {im.get('note','unavailable')}")
-    
-    dd = cm["dependencies"]
-    lines.append(f"- Dep freshness: {dd.get('freshness_percent','?')}% outdated ({dd.get('outdated_deps',0)}/{dd.get('total_deps',0)} deps)")
-    if dd.get('outdated_list'):
-        lines.append(f"  Outdated: {', '.join(dd['outdated_list'][:5])}")
-    
-    content = "\n".join(lines) + "\n"
-    
-    if output_path:
-        output_path.parent.mkdir(parents=True, exist_ok=True)
-        output_path.write_text(content)
-    
-    return content
-
-
-# ── Main ─────────────────────────────────────────────────────────────────────
-
-def main() -> int:
-    parser = argparse.ArgumentParser(description="Progress Tracker — 10.8")
-    parser.add_argument("--json", action="store_true", help="Emit snapshot as JSON only")
-    parser.add_argument("--output", type=Path, default=METRICS_DIR / "TRENDS.md",
-                        help="Write trends markdown to this file")
-    args = parser.parse_args()
-    
-    snapshot = take_snapshot()
-    all_snapshots = load_snapshots()
-    path_written = save_snapshot(snapshot)
-    
-    if args.json:
-        print(json.dumps(snapshot, indent=2))
-        return 0
-    
-    trends = generate_trends(all_snapshots + [snapshot], output_path=args.output)
-    
-    # Print current snapshot summary
-    print(f"Snapshot saved: {path_written}\n")
-    print(f"Progress Tracker — {snapshot['date']}")
-    print("=" * 50)
-    
-    m = snapshot["metrics"]
-    tc = m["test_coverage"]
-    print(f"Test files: {tc['test_files']}  |  Source files: {tc['source_files']}  |  Ratio: {tc['test_to_source_ratio']:.3f}")
-    if tc["coverage_percent"] is not None:
-        print(f"Line coverage: {tc['coverage_percent']:.1f}%  (via {tc['coverage_tool']})")
-    else:
-        print("Line coverage: (not available — run `pytest --cov`)")
-    
-    print()
-    dc = m["doc_coverage"]
-    print(f"Callables with docstrings: {dc['callables_with_doc']}/{dc['callables_total']}  ({dc['doc_coverage_percent']:.1f}%)")
-    
-    print()
-    im = m["issues"]
-    if im.get("close_rate") is not None:
-        print(f"Issues (7d): {im['closed_last_7d']} closed / {im['opened_last_7d']} opened  → close rate: {im['close_rate']*100:.1f}%")
-        print(f"Total open: {im['total_open']}")
-    else:
-        print(f"Issues: {im.get('note','unavailable')}")
-    
-    print()
-    dd = m["dependencies"]
-    print(f"Dependencies: {dd.get('total_deps',0)} total, {dd.get('outdated_deps',0)} outdated")
-    if dd.get('outdated_list'):
-        shown = dd['outdated_list'][:5]
-        print(f"Outdated: {', '.join(shown)}" + ("..." if len(dd['outdated_list']) > 5 else ""))
-    
-    print(f"\nTrends written to: {args.output}")
-    return 0
-
-
-if __name__ == "__main__":
-    sys.exit(main())

scripts/session_pair_harvester.py

@@ -22,114 +22,95 @@ import sys
 from pathlib import Path
 from typing import Optional
 
+from session_reader import extract_conversation, read_session
+
 
 def compute_hash(text: str) -> str:
     """Content hash for deduplication."""
     return hashlib.sha256(text.encode()).hexdigest()[:16]
 
 
-def extract_pairs_from_session(session_data: dict, min_ratio: float = 1.5,
+def extract_pairs_from_conversation(conversation: list, session_id: str, model: str,
+                                min_ratio: float = 1.5,
                                 min_response_words: int = 20) -> list:
-    """Extract terse→rich pairs from a single session object."""
+    """Extract terse→rich pairs from a normalized conversation."""
     pairs = []
-    conversations = session_data.get("conversations", [])
-    session_id = session_data.get("id", "unknown")
-    model = session_data.get("model", "unknown")
-
     seen_hashes = set()
 
-    for i, msg in enumerate(conversations):
-        # Look for assistant/gpt responses
-        if msg.get("from") not in ("gpt", "assistant"):
+    for i, msg in enumerate(conversation):
+        # Look for assistant responses
+        if msg.get('role') != 'assistant':
             continue
 
-        response_text = msg.get("value", "")
+        response_text = msg.get('content', '')
         if not response_text or len(response_text.split()) < min_response_words:
             continue
 
-        # Find the preceding human message
+        # Find the preceding user message
         prompt_text = ""
         for j in range(i - 1, -1, -1):
-            if conversations[j].get("from") == "human":
-                prompt_text = conversations[j].get("value", "")
+            if conversation[j].get('role') == 'user':
+                prompt_text = conversation[j].get('content', '')
                 break
 
         if not prompt_text:
             continue
 
         # Filter: skip tool results, system messages embedded as human
-        if prompt_text.startswith("{") and "output" in prompt_text[:100]:
-            continue  # likely a tool result
-        if prompt_text.startswith("# SOUL.md") or prompt_text.startswith("You are"):
-            continue  # system prompt leak
+        if prompt_text.startswith('{') and 'output' in prompt_text[:100]:
+            continue
+        if prompt_text.startswith('# SOUL.md') or prompt_text.startswith('You are'):
+            continue
 
         # Quality filters
         prompt_words = len(prompt_text.split())
         response_words = len(response_text.split())
 
-        # Must have meaningful length ratio
         if prompt_words == 0 or response_words == 0:
             continue
         ratio = response_words / prompt_words
         if ratio < min_ratio:
             continue
 
-        # Skip responses that are mostly code
-        code_blocks = response_text.count("```")
-        if code_blocks >= 4 and len(response_text.replace("```", "").strip()) < 50:
+        code_blocks = response_text.count('```')
+        if code_blocks >= 4 and len(response_text.replace('```', '').strip()) < 50:
             continue
 
-        # Skip responses with tool call artifacts
-        if "tool_call" in response_text[:100] or "function_call" in response_text[:100]:
+        if 'tool_call' in response_text[:100] or 'function_call' in response_text[:100]:
             continue
 
-        # Deduplicate by content hash
         content_hash = compute_hash(prompt_text + response_text[:200])
         if content_hash in seen_hashes:
             continue
         seen_hashes.add(content_hash)
 
-        # Clean up response: remove markdown headers if too many
         clean_response = response_text
 
         pairs.append({
-            "terse": prompt_text.strip(),
-            "rich": clean_response.strip(),
-            "source": session_id,
-            "model": model,
-            "prompt_words": prompt_words,
-            "response_words": response_words,
-            "ratio": round(ratio, 2),
+            'terse': prompt_text.strip(),
+            'rich': clean_response.strip(),
+            'source': session_id,
+            'model': model,
+            'prompt_words': prompt_words,
+            'response_words': response_words,
+            'ratio': round(ratio, 2),
         })
 
     return pairs
 
 
-def extract_from_jsonl_file(filepath: str, **kwargs) -> list:
-    """Extract pairs from a session JSONL file."""
-    pairs = []
-    path = Path(filepath)
 
-    if not path.exists():
-        print(f"Warning: {filepath} not found", file=sys.stderr)
-        return pairs
-
-    content = path.read_text()
-    lines = content.strip().split("\n")
-
-    for line in lines:
-        line = line.strip()
-        if not line:
-            continue
-        try:
-            session = json.loads(line)
-        except json.JSONDecodeError:
-            continue
-
-        session_pairs = extract_pairs_from_session(session, **kwargs)
-        pairs.extend(session_pairs)
-
-    return pairs
+def extract_from_jsonl_file(path: str, **kwargs) -> list:
+    """Read a session file and extract training pairs using normalized conversation."""
+    session_messages = read_session(path)
+    if not session_messages:
+        return []
+    conversation = extract_conversation(session_messages)
+    # Derive session_id and model from first real message metadata
+    first_msg = next((m for m in session_messages if m.get('role') or m.get('from')), {})
+    session_id = first_msg.get('meta_session_id', Path(path).name)
+    model = first_msg.get('model', 'unknown')
+    return extract_pairs_from_conversation(conversation, session_id, model, **kwargs)
 
 
 def deduplicate_pairs(pairs: list) -> list:

tests/test_session_pair_harvester.py

@@ -0,0 +1,118 @@
+"""
+Tests for session_pair_harvester — training pair extraction from sessions.
+"""
+
+import json
+import tempfile
+import unittest
+from pathlib import Path
+
+import sys
+from pathlib import Path
+sys.path.insert(0, str(Path(__file__).parent.parent / "scripts"))
+from session_pair_harvester import (
+    extract_pairs_from_conversation,
+    extract_from_jsonl_file,
+    deduplicate_pairs,
+    compute_hash,
+)
+
+
+class TestSessionPairHarvester(unittest.TestCase):
+    def test_compute_hash_consistent(self):
+        h1 = compute_hash("hello world")
+        h2 = compute_hash("hello world")
+        self.assertEqual(h1, h2)
+        self.assertEqual(len(h1), 16)
+
+    def test_extract_simple_qa_pair(self):
+        """A simple user→assistant exchange produces one pair."""
+        conversation = [
+            {"role": "user", "content": "What is the capital of France?"},
+            {"role": "assistant", "content": "The capital of France is Paris. It is a major European city renowned for its art, fashion, gastronomy, cultural heritage, and historical significance. The city attracts millions of tourists annually."},
+        ]
+        pairs = extract_pairs_from_conversation(conversation, "test_session", "test-model")
+        self.assertEqual(len(pairs), 1)
+        self.assertEqual(pairs[0]["terse"], "What is the capital of France?")
+        self.assertIn("Paris", pairs[0]["rich"])
+        self.assertEqual(pairs[0]["source"], "test_session")
+
+    def test_min_ratio_filter(self):
+        """Very short responses are filtered out."""
+        conversation = [
+            {"role": "user", "content": "Yes"},
+            {"role": "assistant", "content": "No."},
+        ]
+        # Default min_ratio = 1.5, min_words = 20 for response
+        pairs = extract_pairs_from_conversation(conversation, "s", "m", min_response_words=3)
+        self.assertEqual(len(pairs), 0)
+
+    def test_min_words_filter(self):
+        """Assistant responses below min word count are skipped."""
+        conversation = [
+            {"role": "user", "content": "Explain the project architecture in detail"},
+            {"role": "assistant", "content": "OK."},
+        ]
+        pairs = extract_pairs_from_conversation(conversation, "s", "m", min_response_words=5)
+        self.assertEqual(len(pairs), 0)
+
+    def test_skip_non_assistant_messages(self):
+        """System and tool messages are ignored."""
+        conversation = [
+            {"role": "system", "content": "You are a helpful assistant."},
+            {"role": "user", "content": "Hello"},
+            {"role": "assistant", "content": "Hi there! How can I help you today?"},
+        ]
+        pairs = extract_pairs_from_conversation(conversation, "s", "m", min_response_words=3)
+        self.assertEqual(len(pairs), 1)
+        self.assertEqual(pairs[0]["terse"], "Hello")
+
+    def test_multiple_pairs_from_one_session(self):
+        """A conversation with several Q&A turns yields multiple pairs."""
+        conversation = [
+            {"role": "user", "content": "First question?"},
+            {"role": "assistant", "content": "Here is a detailed and comprehensive answer that thoroughly explores multiple aspects of the subject. It provides background context and practical implications for the reader."},
+            {"role": "user", "content": "Second?"},
+            {"role": "assistant", "content": "Another comprehensive response with detailed examples. This includes practical code blocks and thorough explanations to ensure deep understanding of the topic at hand."},
+        ]
+        pairs = extract_pairs_from_conversation(conversation, "s", "m", min_ratio=1.0)
+        self.assertEqual(len(pairs), 2)
+
+    def test_deduplication_removes_duplicates(self):
+        """Identical pairs across sessions are deduplicated."""
+        pairs = [
+            {"terse": "q1", "rich": "a1", "source": "s1", "model": "m"},
+            {"terse": "q1", "rich": "a1", "source": "s2", "model": "m"},
+            {"terse": "q2", "rich": "a2", "source": "s1", "model": "m"},
+        ]
+        unique = deduplicate_pairs(pairs)
+        self.assertEqual(len(unique), 2)
+        sources = {p["source"] for p in unique}
+        # First unique pair can be from either s1 or s2
+        self.assertIn("s1", sources)
+
+    def test_integration_with_test_sessions(self):
+        """Harvester finds pairs in real test session files."""
+        repo_root = Path(__file__).parent.parent
+        test_sessions_dir = repo_root / "test_sessions"
+        if not test_sessions_dir.exists():
+            self.skipTest("test_sessions not found")
+
+        pairs = []
+        for jsonl_file in sorted(test_sessions_dir.glob("*.jsonl")):
+            pairs.extend(extract_from_jsonl_file(str(jsonl_file)))
+
+        self.assertGreater(len(pairs), 0, "Should extract at least one pair from test_sessions")
+        for p in pairs:
+            self.assertIn("terse", p)
+            self.assertIn("rich", p)
+            self.assertIn("source", p)
+            self.assertIn("model", p)
+            # Verify content exists
+            self.assertGreater(len(p["terse"]), 0)
+            self.assertGreater(len(p["rich"]), 0)
+
+
+if __name__ == "__main__":
+    unittest.main()
+