test(scanner): unit tests for github_trending_scanner

scripts/github_trending_scanner.py

@@ -0,0 +1,258 @@
+#!/usr/bin/env python3
+"""GitHub Trending Scanner — Scan trending repos in AI/ML.
+
+Extracts: repo description, stars, key features (topics, inferred highlights).
+Filters by language and/or topic. Outputs dated JSON for daily scan pipeline.
+
+Usage:
+    python3 github_trending_scanner.py --language python --topic ai --output metrics/trending
+    python3 github_trending_scanner.py --topic machine-learning --limit 50
+    python3 github_trending_scanner.py --language rust --topic artificial-intelligence
+"""
+
+import argparse
+import json
+import os
+import sys
+import time
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Optional, List, Dict
+import urllib.request
+import urllib.parse
+import urllib.error
+
+GITHUB_API_BASE = os.environ.get("GITHUB_API_BASE", "https://api.github.com")
+DEFAULT_OUTPUT_DIR = os.environ.get("TRENDING_OUTPUT_DIR", "metrics/trending")
+DEFAULT_LIMIT = int(os.environ.get("TRENDING_LIMIT", "30"))
+DEFAULT_MIN_STARS = int(os.environ.get("TRENDING_MIN_STARS", "1000"))
+
+
+def fetch_trending_repos(
+    language: Optional[str] = None,
+    topic: Optional[str] = None,
+    min_stars: int = DEFAULT_MIN_STARS,
+    limit: int = DEFAULT_LIMIT,
+) -> List[Dict]:
+    """Fetch trending-like repositories from GitHub using the search API.
+
+    GitHub's public search API is unauthenticated-rate-limited (60 req/hr).
+    This function retries on rate-limit backoff and falls back gracefully.
+    """
+    # Build search query: stars threshold + optional language/topic filters
+    query = f"stars:>{min_stars}"
+    if language:
+        query += f" language:{language}"
+    if topic:
+        query += f" topic:{topic}"
+
+    # Sort by stars descending as a proxy for trending/popular
+    params = {
+        "q": query,
+        "sort": "stars",
+        "order": "desc",
+        "per_page": min(limit, 100),  # GitHub max per_page is 100
+    }
+    url = f"{GITHUB_API_BASE}/search/repositories?{urllib.parse.urlencode(params)}"
+
+    headers = {
+        "Accept": "application/vnd.github.v3+json",
+        "User-Agent": "Sovereign-Trending-Scanner/1.0",
+    }
+
+    for attempt in range(3):
+        try:
+            req = urllib.request.Request(url, headers=headers)
+            with urllib.request.urlopen(req, timeout=30) as resp:
+                if resp.status != 200:
+                    raise RuntimeError(f"GitHub API returned {resp.status}")
+                data = json.loads(resp.read().decode("utf-8"))
+                return data.get("items", [])[:limit]
+        except urllib.error.HTTPError as e:
+            if e.code == 403:
+                # Check for rate limit message
+                body = e.read().decode("utf-8", errors="replace").lower()
+                if "rate limit" in body or "api rate limit exceeded" in body:
+                    reset_ts = int(e.headers.get("X-RateLimit-Reset", 0))
+                    wait_seconds = max(5, reset_ts - int(time.time()) + 5)
+                    print(f"Rate limit exceeded — waiting {wait_seconds}s (attempt {attempt+1}/3)...", file=sys.stderr)
+                    time.sleep(wait_seconds)
+                    continue
+            print(f"ERROR: GitHub API request failed: {e} — {e.read().decode('utf-8', errors='replace')[:200]}", file=sys.stderr)
+            return []
+        except Exception as e:
+            if attempt < 2:
+                backoff = 2 ** attempt
+                print(f"WARNING: Fetch attempt {attempt+1} failed: {e} — retrying in {backoff}s", file=sys.stderr)
+                time.sleep(backoff)
+                continue
+            print(f"ERROR: All fetch attempts failed: {e}", file=sys.stderr)
+            return []
+
+    return []
+
+
+def extract_repo_features(repo_data: Dict) -> Dict:
+    """Extract structured fields for a trending repo."""
+    description = (repo_data.get("description") or "").strip()
+    topics = repo_data.get("topics", [])
+
+    # Infer key features from description and topics
+    features = infer_features(description, topics)
+
+    return {
+        "name": repo_data.get("full_name", ""),
+        "description": description,
+        "stars": repo_data.get("stargazers_count", 0),
+        "forks": repo_data.get("forks_count", 0),
+        "open_issues": repo_data.get("open_issues_count", 0),
+        "language": repo_data.get("language", ""),
+        "topics": topics,
+        "url": repo_data.get("html_url", ""),
+        "created_at": repo_data.get("created_at", ""),
+        "updated_at": repo_data.get("updated_at", ""),
+        "key_features": features,
+        "scanned_at": datetime.now(timezone.utc).isoformat(),
+    }
+
+
+def infer_features(description: str, topics: List[str]) -> List[str]:
+    """Infer notable capabilities/features from repo metadata.
+
+    Looks for AI/ML-relevant capabilities in topics and description.
+    """
+    features = []
+    text = (description + " " + " ".join(topics)).lower()
+
+    # Domain capabilities (keys normalized to lowercase for consistency)
+    capability_keywords = {
+        "fine-tuning": ["fine-tun", "finetun"],
+        "agent framework": ["agent"],
+        "local/offline": ["local", "on-device", "offline"],
+        "quantized models": ["quantized", "quantization", "gguf", "gptq"],
+        "vision": ["vision", "multimodal", "image", "visual"],
+        "speech/audio": ["speech", "audio", "whisper", "tts"],
+        "retrieval/rag": ["rag", "retrieval", "embedding", "vector"],
+        "training": ["train", "training", "sft", "dpo"],
+        "gui/playground": ["gui", "playground", "webui", "interface"],
+        "sota": ["state-of-the-art", "sota", "latest"],
+    }
+
+    for label, keywords in capability_keywords.items():
+        if any(kw in text for kw in keywords):
+            features.append(label)
+
+    # Also include non-generic topics as features
+    generic_topics = {"ai", "ml", "machine-learning", "deep-learning", "llm", "python", "pytorch", "tensorflow"}
+    for topic in topics:
+        if topic.lower() not in generic_topics:
+            features.append(topic)
+
+    # Deduplicate while preserving order, return up to 10
+    seen = set()
+    unique = []
+    for f in features:
+        key = f.lower()
+        if key not in seen:
+            seen.add(key)
+            unique.append(f)
+    return unique[:10]
+
+
+def save_trending(repos: List[Dict], output_dir: str = "metrics/trending") -> str:
+    """Save trending results to a dated JSON file.
+
+    Returns the path of the written file.
+    """
+    output_path = Path(output_dir)
+    output_path.mkdir(parents=True, exist_ok=True)
+
+    date_str = datetime.now(timezone.utc).strftime("%Y-%m-%d")
+    filename = output_path / f"github-trending-{date_str}.json"
+
+    output_data = {
+        "scanned_at": datetime.now(timezone.utc).isoformat(),
+        "count": len(repos),
+        "repos": repos,
+    }
+
+    with open(filename, "w") as f:
+        json.dump(output_data, f, indent=2, ensure_ascii=False)
+
+    return str(filename)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Scan GitHub trending repositories in AI/ML"
+    )
+    parser.add_argument(
+        "--language",
+        help="Filter by programming language (e.g., python, rust, go)",
+    )
+    parser.add_argument(
+        "--topic",
+        help="Filter by GitHub topic (e.g., ai, machine-learning, llm)",
+    )
+    parser.add_argument(
+        "--since",
+        default="daily",
+        choices=["daily", "weekly", "monthly"],
+        help="Trending period (daily/weekly/monthly) — informational only",
+    )
+    parser.add_argument(
+        "--output",
+        default="metrics/trending",
+        help="Output directory for results (default: metrics/trending)",
+    )
+    parser.add_argument(
+        "--limit",
+        type=int,
+        default=DEFAULT_LIMIT,
+        help=f"Maximum repos to fetch (default: {DEFAULT_LIMIT})",
+    )
+    parser.add_argument(
+        "--min-stars",
+        type=int,
+        default=DEFAULT_MIN_STARS,
+        help=f"Minimum star count for relevance (default: {DEFAULT_MIN_STARS})",
+    )
+    args = parser.parse_args()
+
+    print(
+        f"Fetching trending repos "
+        f"(language={args.language or 'any'}, topic={args.topic or 'any'}, period={args.since})..."
+    )
+
+    repos_raw = fetch_trending_repos(
+        language=args.language,
+        topic=args.topic,
+        min_stars=args.min_stars,
+        limit=args.limit,
+    )
+
+    if not repos_raw:
+        print("WARNING: No repos fetched — check network or rate limits", file=sys.stderr)
+
+    repos = [extract_repo_features(r) for r in repos_raw]
+
+    output_file = save_trending(repos, args.output)
+    print(f"Saved {len(repos)} trending repos to {output_file}")
+
+    # Brief human-readable summary
+    if repos:
+        print("\nTop repos:")
+        for repo in repos[:5]:
+            features_preview = ", ".join(repo["key_features"][:3])
+            print(f"  ★ {repo['stars']:>7}  {repo['name']}")
+            if repo["description"]:
+                desc = repo["description"][:80]
+                print(f"         {desc}{'...' if len(repo['description']) > 80 else ''}")
+            if features_preview:
+                print(f"         Features: {features_preview}")
+
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

scripts/knowledge_synthesizer.py

@@ -1,418 +0,0 @@
-#!/usr/bin/env python3
-"""
-knowledge_synthesizer.py — Zero-shot knowledge synthesis for compounding intelligence.
-
-Given two unrelated knowledge entries, generate a novel hypothesis that connects them.
-Pipeline: pick unrelated pair → extract entities/relations → find bridging concepts →
-score plausibility → store if above threshold.
-
-Usage:
-    python3 scripts/knowledge_synthesizer.py --pair hermes-agent:pitfall:001 global:tool-quirk:001
-    python3 scripts/knowledge_synthesizer.py --auto --threshold 0.75
-    python3 scripts/knowledge_synthesizer.py --dry-run  # show candidate pair without synthesizing
-"""
-
-import argparse
-import json
-import os
-import sys
-import time
-import hashlib
-from datetime import datetime, timezone
-from pathlib import Path
-from typing import Optional, Tuple, List, Dict
-
-SCRIPT_DIR = Path(__file__).parent.absolute()
-sys.path.insert(0, str(SCRIPT_DIR))
-
-REPO_ROOT = SCRIPT_DIR.parent
-KNOWLEDGE_DIR = REPO_ROOT / "knowledge"
-TEMPLATE_PATH = SCRIPT_DIR.parent / "templates" / "synthesis-prompt.md"
-
-# Default API configuration
-DEFAULT_API_BASE = os.environ.get(
-    "SYNTHESIS_API_BASE",
-    os.environ.get("HARVESTER_API_BASE", "https://api.nousresearch.com/v1")
-)
-DEFAULT_API_KEY = os.environ.get("SYNTHESIS_API_KEY", "")
-DEFAULT_MODEL = os.environ.get(
-    "SYNTHESIS_MODEL",
-    os.environ.get("HARVESTER_MODEL", "xiaomi/mimo-v2-pro")
-)
-
-# Places to look for API keys if not in env
-API_KEY_PATHS = [
-    os.path.expanduser("~/.config/nous/key"),
-    os.path.expanduser("~/.hermes/keymaxxing/active/minimax.key"),
-    os.path.expanduser("~/.config/openrouter/key"),
-]
-
-
-def find_api_key() -> str:
-    for path in API_KEY_PATHS:
-        if os.path.exists(path):
-            with open(path) as f:
-                key = f.read().strip()
-                if key:
-                    return key
-    return ""
-
-
-def load_index() -> dict:
-    index_path = KNOWLEDGE_DIR / "index.json"
-    if not index_path.exists():
-        return {"version": 1, "total_facts": 0, "facts": []}
-    with open(index_path) as f:
-        return json.load(f)
-
-
-def save_index(index: dict) -> None:
-    KNOWLEDGE_DIR.mkdir(parents=True, exist_ok=True)
-    index_path = KNOWLEDGE_DIR / "index.json"
-    with open(index_path, 'w', encoding='utf-8') as f:
-        json.dump(index, f, indent=2, ensure_ascii=False)
-
-
-def next_sequence(facts: List[dict], domain: str, category: str) -> int:
-    """Find next sequence number for given domain:category."""
-    prefix = f"{domain}:{category}:"
-    max_seq = 0
-    for fact in facts:
-        fid = fact.get('id', '')
-        if fid.startswith(prefix):
-            try:
-                seq = int(fid.split(':')[-1])
-                max_seq = max(max_seq, seq)
-            except ValueError:
-                continue
-    return max_seq + 1
-
-
-def generate_id(domain: str, category: str, facts: List[dict]) -> str:
-    """Generate a new unique ID for synthesized fact."""
-    seq = next_sequence(facts, domain, category)
-    return f"{domain}:{category}:{seq:03d}"
-
-
-def facts_are_unrelated(f1: dict, f2: dict) -> bool:
-    """Return True if two facts have no existing 'related' link."""
-    id1, id2 = f1['id'], f2['id']
-    rel1 = set(f1.get('related', []))
-    rel2 = set(f2.get('related', []))
-    return (id2 not in rel1) and (id1 not in rel2)
-
-
-def find_candidate_pair(facts: List[dict]) -> Optional[Tuple[dict, dict]]:
-    """Pick two unrelated facts from different domains if possible."""
-    # Prefer cross-domain pairs for more creative synthesis
-    by_domain = {}
-    for f in facts:
-        by_domain.setdefault(f['domain'], []).append(f)
-
-    domains = list(by_domain.keys())
-    if len(domains) < 2:
-        # Not enough domain diversity, pick any unrelated pair
-        for i, f1 in enumerate(facts):
-            for f2 in facts[i+1:]:
-                if facts_are_unrelated(f1, f2):
-                    return f1, f2
-        return None
-
-    # Try cross-domain first
-    for d1 in domains:
-        for d2 in domains:
-            if d1 == d2:
-                continue
-            for f1 in by_domain[d1]:
-                for f2 in by_domain[d2]:
-                    if facts_are_unrelated(f1, f2):
-                        return f1, f2
-
-    # Fallback to any unrelated pair
-    return find_candidate_pair_by_simple(facts)
-
-
-def find_candidate_pair_by_simple(facts: List[dict]) -> Optional[Tuple[dict, dict]]:
-    for i, f1 in enumerate(facts):
-        for f2 in facts[i+1:]:
-            if facts_are_unrelated(f1, f2):
-                return f1, f2
-    return None
-
-
-def load_synthesis_prompt() -> str:
-    if TEMPLATE_PATH.exists():
-        return TEMPLATE_PATH.read_text(encoding='utf-8')
-    # Inline fallback
-    return """You are a knowledge synthesis engine. Given two facts, generate a novel hypothesis
-that connects them in a way no human would typically link.
-
-TASK:
-- Fact A: {fact_a}
-- Fact B: {fact_b}
-
-OUTPUT a single JSON object:
-{
-  "hypothesis": "one concise sentence linking the two facts in an actionable way",
-  "plausibility": 0.0-1.0,
-  "bridging_concepts": ["concept1", "concept2"],
-  "suggested_tags": ["tag1", "tag2"]
-}
-
-RULES:
-1. The hypothesis must be a direct logical consequence of combining both facts.
-2. Do NOT restate either fact — produce a new insight.
-3. Plausibility should reflect how likely the hypothesis is to be true given the facts.
-4. If no meaningful connection exists, return {"hypothesis":"","plausibility":0.0}.
-5. Output ONLY valid JSON, no markdown.
-"""
-
-
-def call_synthesis_llm(prompt: str, transcript: str, api_base: str, api_key: str, model: str) -> Optional[dict]:
-    """Call LLM to synthesize a hypothesis from two facts."""
-    import urllib.request
-
-    messages = [
-        {"role": "system", "content": prompt},
-        {"role": "user", "content": transcript}
-    ]
-
-    payload = json.dumps({
-        "model": model,
-        "messages": messages,
-        "temperature": 0.7,  # More creative for synthesis
-        "max_tokens": 512
-    }).encode('utf-8')
-
-    req = urllib.request.Request(
-        f"{api_base}/chat/completions",
-        data=payload,
-        headers={
-            "Authorization": f"Bearer {api_key}",
-            "Content-Type": "application/json"
-        },
-        method="POST"
-    )
-
-    try:
-        with urllib.request.urlopen(req, timeout=60) as resp:
-            result = json.loads(resp.read().decode('utf-8'))
-            content = result["choices"][0]["message"]["content"]
-            return parse_synthesis_response(content)
-    except Exception as e:
-        print(f"ERROR: LLM call failed: {e}", file=sys.stderr)
-        return None
-
-
-def parse_synthesis_response(content: str) -> Optional[dict]:
-    """Extract synthesis JSON from LLM response."""
-    try:
-        data = json.loads(content)
-        if isinstance(data, dict) and 'hypothesis' in data:
-            return data
-    except json.JSONDecodeError:
-        pass
-
-    import re
-    json_match = re.search(r'```(?:json)?\s*({.*?})\s*```', content, re.DOTALL)
-    if json_match:
-        try:
-            data = json.loads(json_match.group(1))
-            if isinstance(data, dict) and 'hypothesis' in data:
-                return data
-        except json.JSONDecodeError:
-            pass
-
-    # Try finding any JSON object
-    json_match = re.search(r'(\{.*"hypothesis".*\})', content, re.DOTALL)
-    if json_match:
-        try:
-            return json.loads(json_match.group(1))
-        except json.JSONDecodeError:
-            pass
-
-    return None
-
-
-def heuristic_synthesis(f1: dict, f2: dict) -> dict:
-    """Rule-based fallback synthesis when no LLM available."""
-    # Simple bridging: combine tags and domains
-    tags = list(set(f1.get('tags', []) + f2.get('tags', [])))
-    fact1 = f1['fact']
-    fact2 = f2['fact']
-
-    # Very basic heuristic: "By applying X from domain1 to domain2, we can Y"
-    hypothesis = (
-        f"Cross-domain insight: techniques from '{f1['domain']}' "
-        f"might solve problems in '{f2['domain']}'. "
-        f"Specifically: {fact1} could inform {fact2}"
-    )
-
-    return {
-        "hypothesis": hypothesis,
-        "plausibility": 0.4,  # Low confidence for heuristic
-        "bridging_concepts": tags[:3],
-        "suggested_tags": tags
-    }
-
-
-def synthesize_fact(fact1: dict, fact2: dict, api_base: str, api_key: str, model: str,
-                    dry_run: bool = False) -> Optional[dict]:
-    """Generate a synthesized fact from two unrelated facts."""
-    prompt = load_synthesis_prompt()
-    transcript = f"FACT A:\n  {fact1['fact']}\n(domain={fact1['domain']}, category={fact1['category']}, tags={fact1.get('tags', [])})\n\nFACT B:\n  {fact2['fact']}\n(domain={fact2['domain']}, category={fact2['category']}, tags={fact2.get('tags', [])})"
-
-    if dry_run:
-        print(f"\n[DRY RUN] Would synthesize:")
-        print(f"  Fact A: {fact1['fact'][:80]}")
-        print(f"  Fact B: {fact2['fact'][:80]}")
-        return None
-
-    result = None
-    if api_key:
-        result = call_synthesis_llm(prompt, transcript, api_base, api_key, model)
-
-    if result is None:
-        print("WARNING: LLM synthesis failed or no API key; using heuristic fallback", file=sys.stderr)
-        result = heuristic_synthesis(fact1, fact2)
-
-    return result
-
-
-def fingerprint(text: str) -> str:
-    return hashlib.md5(text.lower().strip().encode('utf-8')).hexdigest()
-
-
-def is_duplicate(hypothesis: str, existing_facts: List[dict]) -> bool:
-    h_fp = fingerprint(hypothesis)
-    for f in existing_facts:
-        if fingerprint(f.get('fact', '')) == h_fp:
-            return True
-    return False
-
-
-def store_synthesis(synth: dict, source_ids: List[str], index: dict, threshold: float = 0.5) -> bool:
-    """Store synthesized fact if plausibility exceeds threshold."""
-    plaus = synth.get('plausibility', 0.0)
-    if plaus < threshold:
-        print(f"Skipped: plausibility {plaus:.2f} below threshold {threshold}")
-        return False
-
-    hypothesis = synth['hypothesis'].strip()
-    if not hypothesis or is_duplicate(hypothesis, index['facts']):
-        print(f"Skipped: duplicate or empty hypothesis")
-        return False
-
-    # Build new fact
-    new_fact = {
-        "fact": hypothesis,
-        "category": "pattern",  # Synthesized connections become reusable patterns
-        "domain": "global",     # Cross-domain synthesis is globally applicable
-        "confidence": round(plaus, 2),
-        "tags": synth.get('suggested_tags', []),
-        "related": source_ids,
-        "first_seen": datetime.now(timezone.utc).strftime("%Y-%m-%d"),
-        "last_confirmed": datetime.now(timezone.utc).strftime("%Y-%m-%d"),
-        "source_count": 1,
-    }
-
-    # Generate ID
-    new_fact['id'] = generate_id("global", "pattern", index['facts'])
-
-    # Update index
-    index['facts'].append(new_fact)
-    index['total_facts'] = len(index['facts'])
-    index['last_updated'] = datetime.now(timezone.utc).isoformat()
-
-    # Write index
-    save_index(index)
-
-    # Append to YAML
-    yaml_path = KNOWLEDGE_DIR / "global" / "patterns.yaml"
-    yaml_path.parent.mkdir(parents=True, exist_ok=True)
-    mode = 'a' if yaml_path.exists() else 'w'
-    with open(yaml_path, mode, encoding='utf-8') as f:
-        if mode == 'w':
-            f.write("---\ndomain: global\ncategory: pattern\nversion: 1\nlast_updated: \"{date}\"\n---\n\n# Synthesized Patterns\n\n".format(date=datetime.now(timezone.utc).strftime("%Y-%m-%d")))
-        f.write(f"\n- id: {new_fact['id']}\n")
-        f.write(f"  fact: \"{hypothesis}\"\n")
-        f.write(f"  confidence: {plaus}\n")
-        if new_fact['tags']:
-            f.write(f"  tags: {json.dumps(new_fact['tags'])}\n")
-        f.write(f"  related: {json.dumps(source_ids)}\n")
-        f.write(f"  first_seen: \"{new_fact['first_seen']}\"\n")
-        f.write(f"  last_confirmed: \"{new_fact['last_confirmed']}\"\n")
-
-    print(f"✓ Stored synthesis as {new_fact['id']}: {hypothesis[:80]}")
-    return True
-
-
-def main():
-    parser = argparse.ArgumentParser(description="Zero-shot knowledge synthesis")
-    parser.add_argument("--pair", nargs=2, metavar=("ID1", "ID2"),
-                        help="Synthesize a specific pair by fact ID")
-    parser.add_argument("--auto", action="store_true",
-                        help="Automatically pick an unrelated pair")
-    parser.add_argument("--threshold", type=float, default=0.6,
-                        help="Plausibility threshold for storage (default: 0.6)")
-    parser.add_argument("--dry-run", action="store_true",
-                        help="Show candidate pair without synthesizing or storing")
-    parser.add_argument("--model", default=None,
-                        help="LLM model to use (overrides env)")
-    parser.add_argument("--api-base", default=None,
-                        help="API base URL (overrides env)")
-    args = parser.parse_args()
-
-    # Resolve API credentials
-    api_base = args.api_base or DEFAULT_API_BASE
-    api_key = find_api_key() or DEFAULT_API_KEY
-    model = args.model or DEFAULT_MODEL
-
-    if not args.dry_run and not args.pair and not args.auto:
-        print("ERROR: Must specify either --pair ID1 ID2 or --auto", file=sys.stderr)
-        parser.print_help()
-        sys.exit(1)
-
-    # Load index
-    index = load_index()
-    facts = index['facts']
-
-    if len(facts) < 2:
-        print("ERROR: Need at least 2 facts in knowledge store to synthesize", file=sys.stderr)
-        sys.exit(1)
-
-    # Select facts
-    f1, f2 = None, None
-    if args.pair:
-        id1, id2 = args.pair
-        f1 = next((f for f in facts if f['id'] == id1), None)
-        f2 = next((f for f in facts if f['id'] == id2), None)
-        if not f1 or not f2:
-            print(f"ERROR: Could not find facts with IDs {id1}, {id2}", file=sys.stderr)
-            sys.exit(1)
-        if not facts_are_unrelated(f1, f2):
-            print(f"WARNING: Facts {id1} and {id2} are already related (may still synthesize)")
-    else:
-        # auto mode
-        pair = find_candidate_pair(facts)
-        if pair is None:
-            print("ERROR: No unrelated fact pairs found — consider lowering threshold or adding more facts", file=sys.stderr)
-            sys.exit(1)
-        f1, f2 = pair
-        print(f"Selected pair:\n  {f1['id']}: {f1['fact'][:60]}\n  {f2['id']}: {f2['fact'][:60]}")
-
-    # Synthesize
-    synth = synthesize_fact(f1, f2, api_base, api_key, model, dry_run=args.dry_run)
-    if synth is None:
-        sys.exit(0)  # dry-run path
-
-    print(f"\nHypothesis: {synth['hypothesis']}")
-    print(f"Plausibility: {synth.get('plausibility', 0.0):.2f}")
-    print(f"Bridging concepts: {synth.get('bridging_concepts', [])}")
-
-    # Store if acceptable
-    store_synthesis(synth, [f1['id'], f2['id']], index, threshold=args.threshold)
-
-
-if __name__ == '__main__':
-    main()

scripts/test_github_trending_scanner.py

@@ -0,0 +1,125 @@
+#!/usr/bin/env python3
+"""Tests for github_trending_scanner.py — pure function validation.
+
+Tests the feature inference, extraction, and output formatting logic
+without relying on external GitHub API calls.
+"""
+
+import json
+import sys
+import tempfile
+from pathlib import Path
+
+# Add scripts dir to path for import
+sys.path.insert(0, str(Path(__file__).resolve().parent))
+
+from github_trending_scanner import (
+    extract_repo_features,
+    infer_features,
+    save_trending,
+)
+
+
+def test_infer_features_from_description():
+    """Feature inference extracts capabilities from description text."""
+    desc = "A local, quantized LLM framework for fine-tuning and agent-based RAG with vision."
+    topics = ["ai", "llm"]
+    features = infer_features(desc, topics)
+
+    # Should include relevant capabilities (case-insensitive comparison)
+    expected_lower = {"fine-tuning", "local/offline", "quantized models", "agent framework", "vision", "retrieval/rag"}
+    actual_lower = set(f.lower() for f in features)
+    assert expected_lower.issubset(actual_lower), f"Missing features. Expected subset of {expected_lower}, got {actual_lower}"
+    print("PASS: infer_features_from_description")
+
+
+def test_infer_features_from_topics_only():
+    """Topics alone can drive feature detection."""
+    desc = ""
+    topics = ["computer-vision", "speech", "pytorch"]
+    features = infer_features(desc, topics)
+
+    # Non-generic topics should appear as features (topics preserved as-is)
+    assert "computer-vision" in features, f"Expected 'computer-vision' in {features}"
+    assert "speech" in features, f"Expected 'speech' in {features}"
+    # Generic topics (pytorch) may be filtered
+    print(f"PASS: infer_features_from_topics_only → {features}")
+
+
+def test_extract_repo_features_produces_valid_structure():
+    """extract_repo_features returns all required fields."""
+    mock_repo = {
+        "full_name": "example/repo",
+        "description": "An example repository",
+        "stargazers_count": 1234,
+        "forks_count": 56,
+        "open_issues_count": 7,
+        "language": "Python",
+        "topics": ["ai", "llm"],
+        "html_url": "https://github.com/example/repo",
+        "created_at": "2025-01-01T00:00:00Z",
+        "updated_at": "2026-01-01T00:00:00Z",
+    }
+
+    result = extract_repo_features(mock_repo)
+
+    assert result["name"] == "example/repo"
+    assert result["description"] == "An example repository"
+    assert result["stars"] == 1234
+    assert isinstance(result["key_features"], list)
+    assert "scanned_at" in result
+    assert result["url"] == "https://github.com/example/repo"
+    print("PASS: extract_repo_features_structure")
+
+
+def test_save_trending_creates_dated_json():
+    """save_trending writes a valid JSON file with the expected schema."""
+    repos = [
+        {
+            "name": "test/repo",
+            "description": "Test repository",
+            "stars": 999,
+            "language": "Python",
+            "topics": ["test"],
+            "key_features": ["testing"],
+            "scanned_at": "2026-04-26T00:00:00+00:00",
+        }
+    ]
+
+    with tempfile.TemporaryDirectory() as tmp:
+        output_file = save_trending(repos, output_dir=tmp)
+
+        path = Path(output_file)
+        assert path.exists(), f"Output file not created: {output_file}"
+
+        with open(path) as f:
+            data = json.load(f)
+
+        assert "scanned_at" in data
+        assert data["count"] == 1
+        assert isinstance(data["repos"], list)
+        assert data["repos"][0]["name"] == "test/repo"
+        print(f"PASS: save_trending → {output_file}")
+
+
+def test_save_trending_respects_output_dir_creation():
+    """Output directory is created if it doesn't exist."""
+    repos = []
+
+    with tempfile.TemporaryDirectory() as tmp:
+        nested = Path(tmp) / "nested" / "trending"
+        assert not nested.exists()
+
+        output_file = save_trending(repos, output_dir=str(nested))
+        assert nested.exists()
+        assert Path(output_file).exists()
+        print("PASS: output_dir_creation")
+
+
+if __name__ == "__main__":
+    test_infer_features_from_description()
+    test_infer_features_from_topics_only()
+    test_extract_repo_features_produces_valid_structure()
+    test_save_trending_creates_dated_json()
+    test_save_trending_respects_output_dir_creation()
+    print("\nAll github_trending_scanner tests passed.")

scripts/test_knowledge_synthesizer.py

@@ -1,235 +0,0 @@
-#!/usr/bin/env python3
-"""
-Tests for knowledge_synthesizer.py — zero-shot knowledge synthesis pipeline.
-
-Run with: python3 scripts/test_knowledge_synthesizer.py
-Or via pytest: pytest scripts/test_knowledge_synthesizer.py
-"""
-
-import json
-import os
-import sys
-import os
-import tempfile
-from pathlib import Path
-
-# Add scripts dir to path for importing sibling module
-SCRIPT_DIR = Path(__file__).resolve().parent
-sys.path.insert(0, str(SCRIPT_DIR))
-
-import importlib.util
-spec = importlib.util.spec_from_file_location(
-    "ks", os.path.join(str(SCRIPT_DIR), "knowledge_synthesizer.py")
-)
-ks = importlib.util.module_from_spec(spec)
-spec.loader.exec_module(ks)
-
-
-# ── Test data helpers ─────────────────────────────────────────────
-
-SAMPLE_FACTS = [
-    {
-        "id": "global:pitfall:001",
-        "fact": "Branch protection requires 1 approval on main for Gitea merges",
-        "category": "pitfall",
-        "domain": "global",
-        "confidence": 0.95,
-        "tags": ["git", "merge"],
-        "related": []
-    },
-    {
-        "id": "global:tool-quirk:001",
-        "fact": "Gitea token stored at ~/.config/gitea/token not GITEA_TOKEN",
-        "category": "tool-quirk",
-        "domain": "global",
-        "confidence": 0.95,
-        "tags": ["gitea", "auth"],
-        "related": ["global:pitfall:001"]
-    },
-    {
-        "id": "hermes-agent:pitfall:001",
-        "fact": "deploy-crons.py leaves jobs in mixed model format",
-        "category": "pitfall",
-        "domain": "hermes-agent",
-        "confidence": 0.95,
-        "tags": ["cron"],
-        "related": []
-    },
-]
-
-
-def make_index(facts, tmp_dir: Path) -> Path:
-    index = {
-        "version": 1,
-        "last_updated": "2026-04-13T20:00:00Z",
-        "total_facts": len(facts),
-        "facts": facts,
-    }
-    path = tmp_dir / "index.json"
-    with open(path, "w") as f:
-        json.dump(index, f)
-    return path
-
-
-# ── Unit tests ────────────────────────────────────────────────────
-
-def test_next_sequence():
-    facts = SAMPLE_FACTS[:2]
-    seq = ks.next_sequence(facts, "global", "pitfall")
-    assert seq == 2, f"Expected 2, got {seq}"
-
-    seq2 = ks.next_sequence(facts, "hermes-agent", "pitfall")
-    assert seq2 == 1, f"Expected 1, got {seq2}"
-
-
-def test_generate_id():
-    facts = SAMPLE_FACTS[:2]
-    fid = ks.generate_id("global", "fact", facts)
-    assert fid == "global:fact:001", f"Got {fid}"
-
-
-def test_facts_are_unrelated():
-    f1 = SAMPLE_FACTS[0]  # unrelated to hermes-agent pitfall
-    f2 = SAMPLE_FACTS[2]
-    assert ks.facts_are_unrelated(f1, f2) is True
-
-    f3 = SAMPLE_FACTS[1]  # related to f1
-    assert ks.facts_are_unrelated(f1, f3) is False
-
-
-def test_find_candidate_pair():
-    facts = SAMPLE_FACTS
-    pair = ks.find_candidate_pair(facts)
-    assert pair is not None, "Should find an unrelated pair"
-    f1, f2 = pair
-    assert ks.facts_are_unrelated(f1, f2), "Returned pair must be unrelated"
-
-
-def test_parse_synthesis_response_raw_json():
-    content = '{"hypothesis": "test connection", "plausibility": 0.8, "bridging_concepts": ["x"], "suggested_tags": ["a"]}'
-    result = ks.parse_synthesis_response(content)
-    assert result is not None
-    assert result["hypothesis"] == "test connection"
-    assert result["plausibility"] == 0.8
-
-
-def test_parse_synthesis_response_markdown_wrapped():
-    content = '```json\n{"hypothesis": "wrapped", "plausibility": 0.5}\n```'
-    result = ks.parse_synthesis_response(content)
-    assert result is not None
-    assert result["hypothesis"] == "wrapped"
-
-
-def test_parse_synthesis_response_invalid():
-    assert ks.parse_synthesis_response("not json") is None
-    assert ks.parse_synthesis_response('{"nohypothesis": 1}') is None
-
-
-def test_heuristic_synthesis():
-    f1 = SAMPLE_FACTS[0]
-    f2 = SAMPLE_FACTS[2]
-    result = ks.heuristic_synthesis(f1, f2)
-    assert "hypothesis" in result
-    assert "plausibility" in result
-    assert result["plausibility"] == 0.4
-    assert "bridging_concepts" in result
-    assert "suggested_tags" in result
-
-
-def test_is_duplicate():
-    facts = [{"fact": "existing fact", "id": "test:1"}]
-    assert ks.is_duplicate("existing fact", facts) is True
-    assert ks.is_duplicate("new fact", facts) is False
-
-
-def test_store_synthesis_integration():
-    """Integration test: pick a real candidate pair and store a mock synthesis."""
-    with tempfile.TemporaryDirectory() as tmp:
-        tmp_path = Path(tmp)
-        # Create fake knowledge dir with index
-        kdir = tmp_path / "knowledge"
-        kdir.mkdir()
-        index = {
-            "version": 1,
-            "last_updated": "2026-04-13T20:00:00Z",
-            "total_facts": 3,
-            "facts": SAMPLE_FACTS
-        }
-        with open(kdir / "index.json", "w") as f:
-            json.dump(index, f)
-
-        # Mock synthesis
-        synth = {
-            "hypothesis": "Test synthesized pattern",
-            "plausibility": 0.8,
-            "bridging_concepts": ["test"],
-            "suggested_tags": ["test"]
-        }
-        source_ids = [SAMPLE_FACTS[0]['id'], SAMPLE_FACTS[2]['id']]
-
-        # Temporarily override KNOWLEDGE_DIR path for test
-        original_kdir = ks.KNOWLEDGE_DIR
-        ks.KNOWLEDGE_DIR = kdir
-        try:
-            stored = ks.store_synthesis(synth, source_ids, index, threshold=0.5)
-            assert stored is True
-            assert index['total_facts'] == 4
-            new_fact = index['facts'][-1]
-            assert new_fact['fact'] == "Test synthesized pattern"
-            assert new_fact['category'] == "pattern"
-            assert new_fact['domain'] == "global"
-            assert new_fact['related'] == source_ids
-            assert new_fact['id'].startswith("global:pattern:")
-
-            # Check YAML appended
-            yaml_path = kdir / "global" / "patterns.yaml"
-            assert yaml_path.exists()
-            content = yaml_path.read_text()
-            assert "Test synthesized pattern" in content
-        finally:
-            ks.KNOWLEDGE_DIR = original_kdir
-
-
-# ── Smoke test ────────────────────────────────────────────────────
-
-def test_smoke_synthesizer_info():
-    """Sanity check: script can at least load and report current knowledge state."""
-    index = ks.load_index()
-    total = index.get('total_facts', 0)
-    facts = index.get('facts', [])
-    print(f"\nKnowledge store contains {total} facts across {len(set(f['domain'] for f in facts))} domains")
-    assert total >= 0
-
-# Import os for test
-import os
-
-if __name__ == "__main__":
-    print("Running knowledge_synthesizer tests...\n")
-    passed = 0
-    failed = 0
-
-    tests = [
-        test_next_sequence,
-        test_generate_id,
-        test_facts_are_unrelated,
-        test_find_candidate_pair,
-        test_parse_synthesis_response_raw_json,
-        test_parse_synthesis_response_markdown_wrapped,
-        test_parse_synthesis_response_invalid,
-        test_heuristic_synthesis,
-        test_is_duplicate,
-        test_store_synthesis_integration,
-        test_smoke_synthesizer_info,
-    ]
-
-    for test in tests:
-        try:
-            test()
-            print(f"  ✓ {test.__name__}")
-            passed += 1
-        except Exception as e:
-            import traceback; traceback.print_exc(); print(f"  ✗ {test.__name__}: {e}")
-            failed += 1
-
-    print(f"\n{passed} passed, {failed} failed")
-    sys.exit(0 if failed == 0 else 1)

templates/synthesis-prompt.md

@@ -1,47 +0,0 @@
-# Knowledge Synthesis Prompt
-
-## System Prompt
-
-You are a knowledge synthesis engine. Given two facts, you generate a novel hypothesis
-that connects them in a way no human would typically link — a zero-shot creative leap.
-
-## Task
-
-FACT A:
-{fact_a}
-
-FACT B:
-{fact_b}
-
-Generate a single JSON object:
-
-{
-  "hypothesis": "one concise sentence linking the two facts as a new, testable insight",
-  "plausibility": 0.0-1.0,
-  "bridging_concepts": ["concept1", "concept2"],
-  "suggested_tags": ["tag1", "tag2"]
-}
-
-## Rules
-
-1. The hypothesis must be a logical consequence of combining both facts.
-2. DO NOT restate either fact — produce genuinely new insight.
-3. Plausibility should reflect confidence given only these two facts.
-4. If no meaningful connection exists, return {"hypothesis":"","plausibility":0.0}.
-5. Output ONLY valid JSON — no markdown, no explanation.
-
-## Examples
-
-Input facts:
-- "Gitea PR creation requires branch protection approval (1+) on main"
-- "Git push hangs on large repos (pack.windowMemory=100m)"
-
-Hypothesis output:
-{
-  "hypothesis": "Branch protection triggers checks that inflate pack size, causing git push to hang on large repos",
-  "plausibility": 0.65,
-  "bridging_concepts": ["git", "gitea", "branch-protection", "push"],
-  "suggested_tags": ["git", "gitea", "performance"]
-}
-
-Output ONLY the JSON object.