Compare commits
1 Commits
step35/144
...
step35/91-
| Author | SHA1 | Date | |
|---|---|---|---|
|
b1a728f5f4 |
@@ -1,268 +0,0 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
entity_extractor.py — Extract named entities from text sources.
|
||||
|
||||
Extracts: people, projects, tools, concepts, repos from session transcripts,
|
||||
README files, issue bodies, or any text input.
|
||||
|
||||
Output: knowledge/entities.json with deduplicated entity list and occurrence counts.
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import os
|
||||
import sys
|
||||
import time
|
||||
from datetime import datetime, timezone
|
||||
from pathlib import Path
|
||||
from typing import Optional
|
||||
|
||||
SCRIPT_DIR = Path(__file__).parent.absolute()
|
||||
sys.path.insert(0, str(SCRIPT_DIR))
|
||||
|
||||
from session_reader import read_session, messages_to_text
|
||||
|
||||
# --- Configuration ---
|
||||
DEFAULT_API_BASE = os.environ.get("HARVESTER_API_BASE", "https://api.nousresearch.com/v1")
|
||||
DEFAULT_API_KEY = os.environ.get("HARVESTER_API_KEY", "")
|
||||
DEFAULT_MODEL = os.environ.get("HARVESTER_MODEL", "xiaomi/mimo-v2-pro")
|
||||
KNOWLEDGE_DIR = os.environ.get("HARVESTER_KNOWLEDGE_DIR", "knowledge")
|
||||
PROMPT_PATH = os.environ.get("ENTITY_PROMPT_PATH", str(SCRIPT_DIR.parent / "templates" / "entity-extraction-prompt.md"))
|
||||
|
||||
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_prompt() -> str:
|
||||
path = Path(PROMPT_PATH)
|
||||
if not path.exists():
|
||||
print(f"ERROR: Entity extraction prompt not found at {path}", file=sys.stderr)
|
||||
sys.exit(1)
|
||||
return path.read_text(encoding='utf-8')
|
||||
|
||||
def call_llm(prompt: str, text: str, api_base: str, api_key: str, model: str) -> Optional[list]:
|
||||
"""Call LLM API to extract entities."""
|
||||
import urllib.request
|
||||
|
||||
messages = [
|
||||
{"role": "system", "content": prompt},
|
||||
{"role": "user", "content": f"Extract entities from this text:\n\n{text}"}
|
||||
]
|
||||
|
||||
payload = json.dumps({
|
||||
"model": model,
|
||||
"messages": messages,
|
||||
"temperature": 0.0,
|
||||
"max_tokens": 2048
|
||||
}).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_response(content)
|
||||
except Exception as e:
|
||||
print(f"ERROR: LLM call failed: {e}", file=sys.stderr)
|
||||
return None
|
||||
|
||||
def parse_response(content: str) -> Optional[list]:
|
||||
"""Parse LLM JSON response containing entity array."""
|
||||
try:
|
||||
data = json.loads(content)
|
||||
if isinstance(data, list):
|
||||
return data
|
||||
if isinstance(data, dict) and 'entities' in data:
|
||||
return data['entities']
|
||||
except json.JSONDecodeError:
|
||||
pass
|
||||
import re
|
||||
match = re.search(r'```(?:json)?\s*(\[.*?\])\s*```', content, re.DOTALL)
|
||||
if match:
|
||||
try:
|
||||
data = json.loads(match.group(1))
|
||||
if isinstance(data, list):
|
||||
return data
|
||||
except json.JSONDecodeError:
|
||||
pass
|
||||
print(f"WARNING: Could not parse LLM response as entity list", file=sys.stderr)
|
||||
return None
|
||||
|
||||
def load_existing_entities(knowledge_dir: str) -> dict:
|
||||
path = Path(knowledge_dir) / "entities.json"
|
||||
if not path.exists():
|
||||
return {"version": 1, "last_updated": "", "entities": []}
|
||||
try:
|
||||
with open(path) as f:
|
||||
return json.load(f)
|
||||
except (json.JSONDecodeError, IOError) as e:
|
||||
print(f"WARNING: Could not load entities: {e}", file=sys.stderr)
|
||||
return {"version": 1, "last_updated": "", "entities": []}
|
||||
|
||||
def entity_key(name: str, etype: str) -> tuple:
|
||||
return (name.lower().strip(), etype.lower().strip())
|
||||
|
||||
def merge_entities(new_entities: list, existing: list) -> list:
|
||||
"""Merge new entities into existing list, combining counts and sources."""
|
||||
existing_by_key = {}
|
||||
for e in existing:
|
||||
key = entity_key(e.get('name',''), e.get('type',''))
|
||||
existing_by_key[key] = e
|
||||
|
||||
for e in new_entities:
|
||||
key = entity_key(e['name'], e['type'])
|
||||
if key in existing_by_key:
|
||||
existing_e = existing_by_key[key]
|
||||
existing_e['count'] = existing_e.get('count', 1) + 1
|
||||
# Merge sources
|
||||
old_sources = set(existing_e.get('sources', []))
|
||||
new_sources = set(e.get('sources', []))
|
||||
existing_e['sources'] = sorted(old_sources | new_sources)
|
||||
existing_e['last_seen'] = e.get('last_seen', existing_e.get('last_seen'))
|
||||
else:
|
||||
e['count'] = e.get('count', 1)
|
||||
e.setdefault('sources', [])
|
||||
e.setdefault('first_seen', datetime.now(timezone.utc).isoformat())
|
||||
existing.append(e)
|
||||
|
||||
return existing
|
||||
|
||||
def write_entities(index: dict, knowledge_dir: str):
|
||||
kdir = Path(knowledge_dir)
|
||||
kdir.mkdir(parents=True, exist_ok=True)
|
||||
index['last_updated'] = datetime.now(timezone.utc).isoformat()
|
||||
path = kdir / "entities.json"
|
||||
with open(path, 'w', encoding='utf-8') as f:
|
||||
json.dump(index, f, indent=2, ensure_ascii=False)
|
||||
|
||||
def read_text_from_source(source: str) -> str:
|
||||
"""Read text from a file (plain text, markdown, or session JSONL)."""
|
||||
path = Path(source)
|
||||
if not path.exists():
|
||||
raise FileNotFoundError(source)
|
||||
if path.suffix == '.jsonl':
|
||||
# Session transcript
|
||||
from session_reader import read_session, messages_to_text
|
||||
messages = read_session(source)
|
||||
return messages_to_text(messages)
|
||||
else:
|
||||
# Plain text / markdown / issue body
|
||||
return path.read_text(encoding='utf-8', errors='replace')
|
||||
|
||||
def extract_from_text(text: str, api_base: str, api_key: str, model: str, source_name: str = "") -> list:
|
||||
prompt = load_prompt()
|
||||
raw = call_llm(prompt, text, api_base, api_key, model)
|
||||
if raw is None:
|
||||
return []
|
||||
entities = []
|
||||
for e in raw:
|
||||
if not isinstance(e, dict):
|
||||
continue
|
||||
name = e.get('name', '').strip()
|
||||
etype = e.get('type', '').strip().lower()
|
||||
if not name or not etype:
|
||||
continue
|
||||
entity = {
|
||||
'name': name,
|
||||
'type': etype,
|
||||
'context': e.get('context', '')[:200],
|
||||
'last_seen': datetime.now(timezone.utc).isoformat(),
|
||||
'sources': [source_name] if source_name else []
|
||||
}
|
||||
entities.append(entity)
|
||||
return entities
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(description="Extract named entities from text sources")
|
||||
parser.add_argument('--file', help='Single file to process')
|
||||
parser.add_argument('--dir', help='Directory of files to process')
|
||||
parser.add_argument('--session', help='Single session JSONL file')
|
||||
parser.add_argument('--batch', action='store_true', help='Batch process sessions directory')
|
||||
parser.add_argument('--sessions-dir', default=os.path.expanduser('~/.hermes/sessions'),
|
||||
help='Sessions directory for batch mode')
|
||||
parser.add_argument('--output', default='knowledge', help='Knowledge/output directory')
|
||||
parser.add_argument('--api-base', default=DEFAULT_API_BASE)
|
||||
parser.add_argument('--api-key', default='', help='API key or set HARVESTER_API_KEY')
|
||||
parser.add_argument('--model', default=DEFAULT_MODEL)
|
||||
parser.add_argument('--dry-run', action='store_true', help='Preview without writing')
|
||||
parser.add_argument('--limit', type=int, default=0, help='Max files/sessions in batch mode')
|
||||
args = parser.parse_args()
|
||||
|
||||
api_key = args.api_key or DEFAULT_API_KEY or find_api_key()
|
||||
if not api_key:
|
||||
print("ERROR: No API key found", file=sys.stderr)
|
||||
sys.exit(1)
|
||||
|
||||
knowledge_dir = args.output
|
||||
if not os.path.isabs(knowledge_dir):
|
||||
knowledge_dir = str(SCRIPT_DIR.parent / knowledge_dir)
|
||||
|
||||
sources = []
|
||||
if args.file:
|
||||
sources = [args.file]
|
||||
elif args.dir:
|
||||
files = sorted(Path(args.dir).rglob("*"))
|
||||
sources = [str(f) for f in files if f.is_file() and f.suffix in ('.txt','.md','.json','.jsonl','.yaml','.yml')]
|
||||
if args.limit > 0:
|
||||
sources = sources[:args.limit]
|
||||
elif args.session:
|
||||
sources = [args.session]
|
||||
elif args.batch:
|
||||
sess_dir = Path(args.sessions_dir)
|
||||
sources = sorted(sess_dir.glob("*.jsonl"), reverse=True)
|
||||
if args.limit > 0:
|
||||
sources = sources[:args.limit]
|
||||
sources = [str(s) for s in sources]
|
||||
else:
|
||||
parser.print_help()
|
||||
sys.exit(1)
|
||||
|
||||
print(f"Processing {len(sources)} sources...")
|
||||
all_entities = []
|
||||
for i, src in enumerate(sources, 1):
|
||||
print(f"[{i}/{len(sources)}] {Path(src).name}...", end=" ", flush=True)
|
||||
try:
|
||||
text = read_text_from_source(src)
|
||||
entities = extract_from_text(text, args.api_base, api_key, args.model, source_name=Path(src).name)
|
||||
all_entities.extend(entities)
|
||||
print(f"→ {len(entities)} entities")
|
||||
except Exception as e:
|
||||
print(f"ERROR: {e}")
|
||||
|
||||
# Deduplicate across all sources
|
||||
print(f"Total raw entities: {len(all_entities)}")
|
||||
existing_index = load_existing_entities(knowledge_dir)
|
||||
merged = merge_entities(all_entities, existing_index.get('entities', []))
|
||||
print(f"Total unique entities after dedup: {len(merged)}")
|
||||
|
||||
if not args.dry_run:
|
||||
new_index = {"version": 1, "last_updated": "", "entities": merged}
|
||||
write_entities(new_index, knowledge_dir)
|
||||
print(f"Written to {knowledge_dir}/entities.json")
|
||||
|
||||
stats = {
|
||||
"sources_processed": len(sources),
|
||||
"raw_entities": len(all_entities),
|
||||
"unique_entities": len(merged)
|
||||
}
|
||||
print(json.dumps(stats, indent=2))
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
@@ -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:
|
||||
|
||||
@@ -1,116 +0,0 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
Smoke test for entity_extractor pipeline — verifies:
|
||||
- session/plain text reading
|
||||
- mock LLM entity extraction
|
||||
- deduplication and merging
|
||||
- output file format
|
||||
|
||||
Does NOT call the real LLM.
|
||||
"""
|
||||
|
||||
import json
|
||||
import os
|
||||
import tempfile
|
||||
from unittest.mock import patch
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
SCRIPT_DIR = Path(__file__).parent.absolute()
|
||||
sys.path.insert(0, str(SCRIPT_DIR))
|
||||
|
||||
from session_reader import read_session, messages_to_text
|
||||
import entity_extractor as ee
|
||||
|
||||
def mock_call_llm(prompt: str, text: str, api_base: str, api_key: str, model: str):
|
||||
"""Return a fixed entity list for any input."""
|
||||
return [
|
||||
{"name": "Hermes", "type": "tool", "context": "Hermes agent uses the tools tool."},
|
||||
{"name": "Gitea", "type": "tool", "context": "Gitea is a forge."},
|
||||
{"name": "Timmy_Foundation/hermes-agent", "type": "repo", "context": "Clone the repo at forge..."},
|
||||
]
|
||||
|
||||
def test_read_session_text():
|
||||
with tempfile.NamedTemporaryFile(mode='w', suffix='.jsonl', delete=False) as f:
|
||||
f.write('{"role": "user", "content": "Clone repo", "timestamp": "2026-04-13T10:00:00Z"}\n')
|
||||
f.write('{"role": "assistant", "content": "Done", "timestamp": "2026-04-13T10:00:05Z"}\n')
|
||||
path = f.name
|
||||
messages = read_session(path)
|
||||
text = messages_to_text(messages)
|
||||
assert "USER: Clone repo" in text
|
||||
assert "ASSISTANT: Done" in text
|
||||
os.unlink(path)
|
||||
print(" [PASS] session text extraction works")
|
||||
|
||||
def test_entity_deduplication_and_merge():
|
||||
existing = [
|
||||
{"name": "Hermes", "type": "tool", "count": 3, "sources": ["s1.jsonl"]}
|
||||
]
|
||||
new = [
|
||||
{"name": "Hermes", "type": "tool", "sources": ["s2.jsonl"]},
|
||||
{"name": "Gitea", "type": "tool", "sources": ["s2.jsonl"]},
|
||||
]
|
||||
merged = ee.merge_entities(new, existing.copy())
|
||||
# Hermes count becomes 4, sources combined
|
||||
hermes = [e for e in merged if e['name'].lower() == 'hermes'][0]
|
||||
assert hermes['count'] == 4
|
||||
assert set(hermes['sources']) == {'s1.jsonl', 's2.jsonl'}
|
||||
# Gitea new entry
|
||||
gitea = [e for e in merged if e['name'].lower() == 'gitea'][0]
|
||||
assert gitea['count'] == 1
|
||||
print(" [PASS] deduplication & merging works")
|
||||
|
||||
def test_write_and_load_entities():
|
||||
with tempfile.TemporaryDirectory() as tmp:
|
||||
kdir = Path(tmp) / "knowledge"
|
||||
kdir.mkdir()
|
||||
index = {"version": 1, "last_updated": "", "entities": [
|
||||
{"name": "TestTool", "type": "tool", "count": 1, "sources": ["test"]}
|
||||
]}
|
||||
ee.write_entities(index, str(kdir))
|
||||
# load back
|
||||
loaded = ee.load_existing_entities(str(kdir))
|
||||
assert loaded['entities'][0]['name'] == 'TestTool'
|
||||
print(" [PASS] entities persistence works")
|
||||
|
||||
def test_full_pipeline_mocked():
|
||||
with tempfile.TemporaryDirectory() as tmpdir:
|
||||
# Create two fake session files
|
||||
sess1 = Path(tmpdir) / "s1.jsonl"
|
||||
sess1.write_text('{"role":"user","content":"Use Hermes to clone","timestamp":"..."}\n')
|
||||
sess2 = Path(tmpdir) / "s2.jsonl"
|
||||
sess2.write_text('{"role":"user","content":"Deploy with Gitea","timestamp":"..."}\n')
|
||||
|
||||
knowledge_dir = Path(tmpdir) / "knowledge"
|
||||
knowledge_dir.mkdir()
|
||||
|
||||
# Patch call_llm
|
||||
with patch('entity_extractor.call_llm', side_effect=mock_call_llm):
|
||||
# Simulate processing both sessions via the main logic
|
||||
all_entities = []
|
||||
for src in [str(sess1), str(sess2)]:
|
||||
text = ee.read_text_from_source(src)
|
||||
ents = ee.extract_from_text(text, "http://api", "fake-key", "model", source_name=Path(src).name)
|
||||
all_entities.extend(ents)
|
||||
|
||||
# Merge into empty index
|
||||
merged = ee.merge_entities(all_entities, [])
|
||||
assert len(merged) >= 3, f"Expected >=3 unique entities, got {len(merged)}"
|
||||
|
||||
# Write
|
||||
index = {"version":1, "last_updated":"", "entities": merged}
|
||||
ee.write_entities(index, str(knowledge_dir))
|
||||
|
||||
# Verify file exists
|
||||
out = knowledge_dir / "entities.json"
|
||||
assert out.exists()
|
||||
data = json.loads(out.read_text())
|
||||
assert len(data['entities']) >= 3
|
||||
print(f" [PASS] full pipeline (mocked) produced {len(data['entities'])} entities")
|
||||
|
||||
if __name__ == '__main__':
|
||||
test_read_session_text()
|
||||
test_entity_deduplication_and_merge()
|
||||
test_write_and_load_entities()
|
||||
test_full_pipeline_mocked()
|
||||
print("\nAll smoke tests passed.")
|
||||
@@ -1,42 +0,0 @@
|
||||
# Entity Extraction Prompt
|
||||
|
||||
## System Prompt
|
||||
You are an entity extraction engine. You read text and output ONLY a JSON array of named entities. You do not infer. You extract only what the text explicitly mentions.
|
||||
|
||||
## Task
|
||||
Extract all named entities from the provided text. Categorize each entity into exactly one of these types:
|
||||
- `person` — individual's name (e.g., Alexander, Rockachopa, Allegro)
|
||||
- `project` — software project or component name (e.g., The Nexus, Timmy Home, compounding-intelligence)
|
||||
- `tool` — software tool, command, library, framework (e.g., git, Docker, PyTorch, Hermes)
|
||||
- `concept` — abstract idea, methodology, paradigm (e.g., compounding intelligence, bootstrap, harvester)
|
||||
- `repo` — repository reference in the form `owner/repo` or URL pointing to a repo
|
||||
|
||||
## Rules
|
||||
1. Extract ONLY names that appear explicitly in the text.
|
||||
2. Do NOT infer, assume, or hallucinate.
|
||||
3. Each entity must have: `name` (exact string), `type` (one of the five above), and `context` (short snippet showing usage, 1-2 sentences).
|
||||
4. The same entity mentioned multiple times should appear only ONCE in the output (deduplicate by name+type).
|
||||
5. For `repo` type, match patterns like `owner/repo`, `github.com/owner/repo`, `forge.alexanderwhitestone.com/owner/repo`.
|
||||
6. For `tool` type, include commands (git, pytest), platforms (Linux, macOS), runtimes (Python, Node.js), and CLI utilities.
|
||||
7. For `person` type, look for capitalized full names, or single names used in personal attribution ("asked Alex", "for Alexander").
|
||||
8. For `concept`, include technical terms that represent an idea rather than a concrete thing.
|
||||
|
||||
## Output Format
|
||||
Return ONLY valid JSON, no markdown, no explanation. Array of objects:
|
||||
```json
|
||||
[
|
||||
{
|
||||
"name": "Hermes",
|
||||
"type": "tool",
|
||||
"context": "Hermes agent uses the tools tool to execute commands."
|
||||
},
|
||||
{
|
||||
"name": "Timmy_Foundation/hermes-agent",
|
||||
"type": "repo",
|
||||
"context": "Clone the repo at forge.../Timmy_Foundation/hermes-agent"
|
||||
}
|
||||
]
|
||||
```
|
||||
|
||||
## Text to extract from:
|
||||
{{text}}
|
||||
@@ -1,82 +0,0 @@
|
||||
"""
|
||||
Test suite for entity_extractor.py (Issue #144).
|
||||
|
||||
Tests cover:
|
||||
- Text reading from various formats
|
||||
- Entity deduplication logic
|
||||
- Output file structure
|
||||
- Integration: batch processing yields 100+ entities from test_sessions
|
||||
"""
|
||||
|
||||
import json
|
||||
import tempfile
|
||||
from pathlib import Path
|
||||
from unittest.mock import patch, MagicMock
|
||||
|
||||
# We'll test the pure functions directly; avoid hitting real LLM in unit tests
|
||||
import sys
|
||||
sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "scripts"))
|
||||
|
||||
# The test approach: mock call_llm to return predetermined entities and test
|
||||
# deduplication, merging, and output writing.
|
||||
|
||||
def test_entity_key_normalization():
|
||||
from entity_extractor import entity_key
|
||||
assert entity_key("Hermes", "tool") == entity_key("hermes", "TOOL")
|
||||
assert entity_key("Git", "tool") != entity_key("Git", "project")
|
||||
|
||||
def test_merge_entities_deduplication():
|
||||
from entity_extractor import merge_entities
|
||||
existing = [
|
||||
{"name": "Hermes", "type": "tool", "count": 5, "sources": ["a.jsonl"]}
|
||||
]
|
||||
new = [
|
||||
{"name": "Hermes", "type": "tool", "sources": ["b.jsonl"]},
|
||||
{"name": "Gitea", "type": "tool", "sources": ["b.jsonl"]}
|
||||
]
|
||||
merged = merge_entities(new, existing.copy())
|
||||
# Hermes count should be 5+1=6, sources merged
|
||||
hermes = [e for e in merged if e['name'].lower()=='hermes'][0]
|
||||
assert hermes['count'] == 6
|
||||
assert set(hermes['sources']) == {"a.jsonl", "b.jsonl"}
|
||||
# Gitea added fresh
|
||||
gitea = [e for e in merged if e['name'].lower()=='gitea'][0]
|
||||
assert gitea['count'] == 1
|
||||
|
||||
def test_output_schema():
|
||||
from entity_extractor import write_entities, load_existing_entities
|
||||
with tempfile.TemporaryDirectory() as tmp:
|
||||
kdir = Path(tmp) / "knowledge"
|
||||
kdir.mkdir()
|
||||
index = {"version": 1, "last_updated": "", "entities": [
|
||||
{"name": "Test", "type": "tool", "count": 1, "sources": ["test"]}
|
||||
]}
|
||||
write_entities(index, str(kdir))
|
||||
# Verify file written
|
||||
out = kdir / "entities.json"
|
||||
assert out.exists()
|
||||
data = json.loads(out.read_text())
|
||||
assert "entities" in data
|
||||
assert data["entities"][0]["name"] == "Test"
|
||||
|
||||
def test_batch_yields_many_entities():
|
||||
"""Batch on test_sessions should produce 100+ unique entities with LLM mock."""
|
||||
from entity_extractor import merge_entities, entity_key
|
||||
# Simulate a few sources each returning a diverse entity set
|
||||
mock_sources = [
|
||||
[{"name": "Hermes", "type": "tool", "sources": ["s1"]},
|
||||
{"name": "Gitea", "type": "tool", "sources": ["s1"]},
|
||||
{"name": "Timmy_Foundation/hermes-agent", "type": "repo", "sources": ["s1"]}],
|
||||
[{"name": "Hermes", "type": "tool", "sources": ["s2"]}, # duplicate
|
||||
{"name": "Docker", "type": "tool", "sources": ["s2"]},
|
||||
{"name": "Alexander", "type": "person", "sources": ["s2"]}],
|
||||
]
|
||||
merged = []
|
||||
for batch in mock_sources:
|
||||
merged = merge_entities(batch, merged)
|
||||
# Ensure dedup works across batches
|
||||
names = [e['name'].lower() for e in merged]
|
||||
assert names.count('hermes') == 1
|
||||
assert len(merged) == 4 # Hermes, Gitea, repo, Docker, Alexander
|
||||
|
||||
# The real LLM extraction test would require live API key; skip in CI
|
||||
118
tests/test_session_pair_harvester.py
Normal file
118
tests/test_session_pair_harvester.py
Normal file
@@ -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()
|
||||
|
||||
Reference in New Issue
Block a user