Compare commits

..

1 Commits

Author SHA1 Message Date
cc215e3ed7 feat: knowledge deduplication — content hash + token similarity (#196)
Some checks failed
Test / pytest (pull_request) Failing after 21s
Dedup module for knowledge entries with:
- SHA256 content hashing for exact duplicates
- Token Jaccard similarity for near-duplicates (default 0.95)
- Quality-based merge: keeps higher confidence/source_count
- Metadata merging: tags, related, source_count
- Dry-run mode
- 30 tests passing
- Built-in --test mode with generated duplicates

Usage:
  python scripts/dedup.py --input knowledge/index.json
  python scripts/dedup.py --input knowledge/index.json --dry-run
  python scripts/dedup.py --test

Closes #196.
2026-04-21 07:58:09 -04:00
3 changed files with 724 additions and 212 deletions

412
GENOME.md
View File

@@ -1,16 +1,16 @@
# GENOME.md — compounding-intelligence
**Generated:** 2026-04-17
**Repo:** Timmy_Foundation/compounding-intelligence
**Description:** Turn 1B+ daily agent tokens into durable, compounding fleet intelligence.
*Auto-generated codebase genome. Addresses timmy-home#676.*
---
## Project Overview
Every agent session starts at zero. The same HTTP 405 gets rediscovered as a branch protection issue. The same token path gets searched from scratch. Intelligence evaporates when the session ends.
**What:** A system that turns 1B+ daily agent tokens into durable, compounding fleet intelligence.
Compounding-intelligence solves this with three pipelines forming a loop:
**Why:** Every agent session starts at zero. The same mistakes get made repeatedly — the same HTTP 405 is rediscovered as a branch protection issue, the same token path is searched for from scratch. Intelligence evaporates when the session ends.
**How:** Three pipelines form a compounding loop:
```
SESSION ENDS → HARVESTER → KNOWLEDGE STORE → BOOTSTRAPPER → NEW SESSION STARTS SMARTER
@@ -18,234 +18,222 @@ SESSION ENDS → HARVESTER → KNOWLEDGE STORE → BOOTSTRAPPER → NEW SESSION
MEASURER → Prove it's working
```
**Status:** Active development. Core pipelines implemented. 20+ scripts, 14 test files, knowledge store populated with real data.
**Status:** Early stage. Template and test scaffolding exist. Core pipeline scripts (harvester.py, bootstrapper.py, measurer.py, session_reader.py) are planned but not yet implemented. The knowledge extraction prompt is complete and validated.
---
## Architecture
```mermaid
graph TD
TRANS[Session Transcripts<br/>~/.hermes/sessions/*.jsonl] --> READER[session_reader.py]
READER --> HARVESTER[harvester.py]
HARVESTER -->|LLM extraction| PROMPT[harvest-prompt.md]
HARVESTER --> DEDUP[deduplicate()]
DEDUP --> INDEX[knowledge/index.json]
DEDUP --> GLOBAL[knowledge/global/*.yaml]
DEDUP --> REPO[knowledge/repos/*.yaml]
INDEX --> BOOTSTRAPPER[bootstrapper.py]
BOOTSTRAPPER -->|filter + rank + truncate| CONTEXT[Bootstrap Context<br/>2k token injection]
CONTEXT --> SESSION[New Session starts smarter]
INDEX --> VALIDATOR[validate_knowledge.py]
INDEX --> STALENESS[knowledge_staleness_check.py]
INDEX --> GAPS[knowledge_gap_identifier.py]
TRANS --> SAMPLER[sampler.py]
SAMPLER -->|score + rank| BEST[High-value sessions]
BEST --> HARVESTER
TRANS --> METADATA[session_metadata.py]
METADATA --> SUMMARY[SessionSummary objects]
KNOWLEDGE --> DIFF[diff_analyzer.py]
DIFF --> PROPOSALS[improvement_proposals.py]
PROPOSALS --> PRIORITIES[priority_rebalancer.py]
A[Session Transcript<br/>.jsonl] --> B[Harvester]
B --> C{Extract Knowledge}
C --> D[knowledge/index.json]
C --> E[knowledge/global/*.md]
C --> F[knowledge/repos/{repo}.md]
C --> G[knowledge/agents/{agent}.md]
D --> H[Bootstrapper]
H --> I[Bootstrap Context<br/>2k token injection]
I --> J[New Session<br/>starts smarter]
J --> A
D --> K[Measurer]
K --> L[metrics/dashboard.md]
K --> M[Velocity / Hit Rate<br/>Error Reduction]
```
## Entry Points
### Pipeline 1: Harvester
### Core Pipelines
**Status:** Prompt designed. Script not implemented.
| Script | Purpose | Key Functions |
|--------|---------|---------------|
| `harvester.py` | Extract knowledge from session transcripts | `harvest_session()`, `call_llm()`, `deduplicate()`, `validate_fact()` |
| `bootstrapper.py` | Build pre-session context from knowledge store | `build_bootstrap_context()`, `filter_facts()`, `sort_facts()`, `truncate_to_tokens()` |
| `session_reader.py` | Parse JSONL session transcripts | `read_session()`, `extract_conversation()`, `messages_to_text()` |
| `sampler.py` | Score and rank sessions for harvesting value | `scan_session_fast()`, `score_session()` |
| `session_metadata.py` | Extract structured metadata from sessions | `extract_session_metadata()`, `SessionSummary` |
Reads finished session transcripts (JSONL). Uses `templates/harvest-prompt.md` to extract durable knowledge into five categories:
### Analysis & Quality
| Category | Description | Example |
|----------|-------------|---------|
| `fact` | Concrete, verifiable information | "Repository X has 5 files" |
| `pitfall` | Errors encountered, wrong assumptions | "Token is at ~/.config/gitea/token, not env var" |
| `pattern` | Successful action sequences | "Deploy: test → build → push → webhook" |
| `tool-quirk` | Environment-specific behaviors | "URL format requires trailing slash" |
| `question` | Identified but unanswered | "Need optimal batch size for harvesting" |
| Script | Purpose |
|--------|---------|
| `validate_knowledge.py` | Validate knowledge index schema compliance |
| `knowledge_staleness_check.py` | Detect stale knowledge (source changed since extraction) |
| `knowledge_gap_identifier.py` | Find untested functions, undocumented APIs, missing tests |
| `diff_analyzer.py` | Analyze code diffs for improvement signals |
| `improvement_proposals.py` | Generate ranked improvement proposals |
| `priority_rebalancer.py` | Rebalance priorities across proposals |
| `automation_opportunity_finder.py` | Find manual steps that can be automated |
| `dead_code_detector.py` | Detect unused code |
| `dependency_graph.py` | Map dependency relationships |
| `perf_bottleneck_finder.py` | Find performance bottlenecks |
| `refactoring_opportunity_finder.py` | Identify refactoring targets |
| `gitea_issue_parser.py` | Parse Gitea issues for knowledge extraction |
### Automation
| Script | Purpose |
|--------|---------|
| `session_pair_harvester.py` | Extract training pairs from sessions |
## Data Flow
```
1. Session ends → .jsonl written to ~/.hermes/sessions/
2. sampler.py scores sessions by age, recency, repo coverage
3. harvester.py reads top sessions, calls LLM with harvest-prompt.md
4. LLM extracts facts/pitfalls/patterns/quirks/questions
5. deduplicate() checks against existing index via fact_fingerprint()
6. validate_fact() checks schema compliance
7. write_knowledge() appends to knowledge/index.json + per-repo YAML
8. On next session start, bootstrapper.py:
a. Loads knowledge/index.json
b. Filters by session's repo and agent type
c. Sorts by confidence (high first), then recency
d. Truncates to 2k token budget
e. Injects as pre-context
9. Agent starts with full situational awareness instead of zero
```
## Key Abstractions
### Knowledge Item (fact/pitfall/pattern/quirk/question)
Output schema per knowledge item:
```json
{
"fact": "Gitea token is at ~/.config/gitea/token",
"category": "tool-quirk",
"repo": "global",
"confidence": 0.9,
"evidence": "Found during clone attempt",
"source_session": "2026-04-13_abc123",
"extracted_at": "2026-04-13T20:00:00Z"
"fact": "One sentence description",
"category": "fact|pitfall|pattern|tool-quirk|question",
"repo": "repo-name or 'global'",
"confidence": 0.0-1.0
}
```
### SessionSummary (session_metadata.py)
Extracted metadata per session: duration, token count, tools used, repos touched, error count, outcome.
### Pipeline 2: Bootstrapper
### Gap / GapReport (knowledge_gap_identifier.py)
Structured gap analysis: untested functions, undocumented APIs, missing tests. Severity: critical/high/medium/low.
**Status:** Not implemented.
### Knowledge Index (knowledge/index.json)
Machine-readable fact store. 12KB, populated with real data. Categories: fact, pitfall, pattern, tool-quirk, question.
Queries knowledge store before session start. Assembles a compact 2k-token context from relevant facts. Injects into session startup so the agent begins with full situational awareness.
## Knowledge Store
### Pipeline 3: Measurer
```
knowledge/
├── index.json # Master fact store (12KB, populated)
├── SCHEMA.md # Schema documentation
├── global/
│ ├── pitfalls.yaml # Cross-repo pitfalls (2KB)
│ └── tool-quirks.yaml # Tool-specific quirks (2KB)
├── repos/
│ ├── hermes-agent.yaml # hermes-agent knowledge (2KB)
│ └── the-nexus.yaml # the-nexus knowledge (2KB)
└── agents/ # Per-agent knowledge (empty)
```
**Status:** Not implemented.
## API Surface
### LLM API (consumed)
| Provider | Endpoint | Usage |
|----------|----------|-------|
| Nous Research | `https://inference-api.nousresearch.com/v1` | Knowledge extraction |
| Ollama | `http://localhost:11434/v1` | Local fallback |
### File API (consumed/produced)
| Path | Format | Direction |
|------|--------|-----------|
| `~/.hermes/sessions/*.jsonl` | JSONL | Input (session transcripts) |
| `knowledge/index.json` | JSON | Output (master fact store) |
| `knowledge/global/*.yaml` | YAML | Output (cross-repo knowledge) |
| `knowledge/repos/*.yaml` | YAML | Output (per-repo knowledge) |
| `templates/harvest-prompt.md` | Markdown | Config (extraction prompt) |
## Test Coverage
**14 test files** covering core pipelines:
| Test File | Covers |
|-----------|--------|
| `test_harvest_prompt.py` | Prompt validation, hallucination detection |
| `test_harvest_prompt_comprehensive.py` | Extended prompt testing |
| `test_harvester_pipeline.py` | Harvester extraction + dedup |
| `test_bootstrapper.py` | Context building, filtering, truncation |
| `test_session_pair_harvester.py` | Training pair extraction |
| `test_improvement_proposals.py` | Proposal generation |
| `test_priority_rebalancer.py` | Priority scoring |
| `test_knowledge_staleness.py` | Staleness detection |
| `test_automation_opportunity_finder.py` | Automation detection |
| `test_diff_analyzer.py` | Diff analysis |
| `test_gitea_issue_parser.py` | Issue parsing |
| `test_refactoring_opportunity_finder.py` | Refactoring signals |
| `test_knowledge_gap_identifier.py` | Gap analysis |
| `test_perf_bottleneck_finder.py` | Perf bottleneck detection |
### Coverage Gaps
1. **session_reader.py** — No dedicated test file (tested indirectly)
2. **sampler.py** — No test file (scoring logic untested)
3. **session_metadata.py** — No test file
4. **validate_knowledge.py** — No test file
5. **knowledge_staleness_check.py** — Tested but limited
## Security Considerations
### API Key Handling
- `harvester.py` reads API key from `~/.hermes/auth.json` or env vars
- Key passed to LLM API in request headers only
- No key logging
### Knowledge Integrity
- `validate_fact()` checks schema before writing
- `deduplicate()` prevents duplicate entries via fingerprint
- `knowledge_staleness_check.py` detects when source code changed but knowledge didn't
- Confidence scores prevent low-quality knowledge from polluting the store
### File Safety
- Knowledge writes are append-only (never deletes)
- Bootstrap context is truncated to budget (no prompt injection via knowledge)
- Session reader handles malformed JSONL gracefully
## File Index
```
scripts/
harvester.py (473 lines) — Core knowledge extraction
bootstrapper.py (302 lines) — Pre-session context builder
session_reader.py (137 lines) — JSONL session parser
sampler.py (363 lines) — Session scoring + ranking
session_metadata.py (271 lines) — Session metadata extraction
validate_knowledge.py (44 lines) — Index validation
knowledge_staleness_check.py (125 lines) — Staleness detection
knowledge_gap_identifier.py (291 lines) — Gap analysis engine
diff_analyzer.py (203 lines) — Diff analysis
improvement_proposals.py (518 lines) — Proposal generation
priority_rebalancer.py (745 lines) — Priority scoring
automation_opportunity_finder.py (600 lines) — Automation detection
dead_code_detector.py (270 lines) — Dead code detection
dependency_graph.py (220 lines) — Dependency mapping
perf_bottleneck_finder.py (635 lines) — Perf analysis
refactoring_opportunity_finder.py (46 lines) — Refactoring signals
gitea_issue_parser.py (140 lines) — Gitea issue parsing
session_pair_harvester.py (224 lines) — Training pair extraction
knowledge/
index.json (12KB) — Master fact store
SCHEMA.md (3KB) — Schema docs
global/pitfalls.yaml (2KB) — Cross-repo pitfalls
global/tool-quirks.yaml (2KB) — Tool quirks
repos/hermes-agent.yaml (2KB) — Repo-specific knowledge
repos/the-nexus.yaml (2KB) — Repo-specific knowledge
templates/
harvest-prompt.md (4KB) — Extraction prompt
test_sessions/ (5 files) — Sample transcripts
tests/ + scripts/test_* (14 files)— Test suite
```
**Total:** ~6,500 lines of code across 18 scripts + 14 test files.
Tracks compounding metrics: knowledge velocity (facts/day), error reduction (%), hit rate (knowledge used / knowledge available), task completion improvement.
---
*Generated by Codebase Genome pipeline — Issue #676*
## Directory Structure
```
compounding-intelligence/
├── README.md # Project overview and architecture
├── GENOME.md # This file (codebase genome)
├── knowledge/ # [PLANNED] Knowledge store
│ ├── index.json # Machine-readable fact index
│ ├── global/ # Cross-repo knowledge
│ ├── repos/{repo}.md # Per-repo knowledge
│ └── agents/{agent}.md # Agent-type notes
├── scripts/
│ ├── test_harvest_prompt.py # Basic prompt validation (2.5KB)
│ └── test_harvest_prompt_comprehensive.py # Full prompt structure test (6.8KB)
├── templates/
│ └── harvest-prompt.md # Knowledge extraction prompt (3.5KB)
├── test_sessions/
│ ├── session_success.jsonl # Happy path test data
│ ├── session_failure.jsonl # Failure path test data
│ ├── session_partial.jsonl # Incomplete session test data
│ ├── session_patterns.jsonl # Pattern extraction test data
│ └── session_questions.jsonl # Question identification test data
└── metrics/ # [PLANNED] Compounding metrics
└── dashboard.md
```
---
## Entry Points and Data Flow
### Entry Point 1: Knowledge Extraction (Harvester)
```
Input: Session transcript (JSONL)
templates/harvest-prompt.md (LLM prompt)
Knowledge items (JSON array)
Output: knowledge/index.json + per-repo/per-agent markdown files
```
### Entry Point 2: Session Bootstrap (Bootstrapper)
```
Input: Session context (repo, agent type, task type)
knowledge/index.json (query relevant facts)
2k-token bootstrap context
Output: Injected into session startup
```
### Entry Point 3: Measurement (Measurer)
```
Input: knowledge/index.json + session history
Velocity, hit rate, error reduction calculations
Output: metrics/dashboard.md
```
---
## Key Abstractions
### Knowledge Item
The atomic unit. One sentence, one category, one confidence score. Designed to be small enough that 1000 items fit in a 2k-token bootstrap context.
### Knowledge Store
A directory structure that mirrors the fleet's mental model:
- `global/` — knowledge that applies everywhere (tool quirks, environment facts)
- `repos/` — knowledge specific to each repo
- `agents/` — knowledge specific to each agent type
### Confidence Score
0.01.0 scale. Defines how certain the harvester is about each extracted fact:
- 0.91.0: Explicitly stated with verification
- 0.70.8: Clearly implied by multiple data points
- 0.50.6: Suggested but not fully verified
- 0.30.4: Inferred from limited data
- 0.10.2: Speculative or uncertain
### Bootstrap Context
The 2k-token injection that a new session receives. Assembled from the most relevant knowledge items for the current task, filtered by confidence > 0.7, deduplicated, and compressed.
---
## API Surface
### Internal (scripts not yet implemented)
| Script | Input | Output | Status |
|--------|-------|--------|--------|
| `harvester.py` | Session JSONL path | Knowledge items JSON | PLANNED |
| `bootstrapper.py` | Repo + agent type | 2k-token context string | PLANNED |
| `measurer.py` | Knowledge store path | Metrics JSON | PLANNED |
| `session_reader.py` | Session JSONL path | Parsed transcript | PLANNED |
### Prompt (templates/harvest-prompt.md)
The extraction prompt is the core "API." It takes a session transcript and returns structured JSON. It defines:
- Five extraction categories
- Output format (JSON array of knowledge items)
- Confidence scoring rubric
- Constraints (no hallucination, specificity, relevance, brevity)
- Example input/output pair
---
## Test Coverage
### What Exists
| File | Tests | Coverage |
|------|-------|----------|
| `scripts/test_harvest_prompt.py` | 2 tests | Prompt file existence, sample transcript |
| `scripts/test_harvest_prompt_comprehensive.py` | 5 tests | Prompt structure, categories, fields, confidence scoring, size limits |
| `test_sessions/*.jsonl` | 5 sessions | Success, failure, partial, patterns, questions |
### What's Missing
1. **Harvester integration test** — Does the prompt actually extract correct knowledge from real transcripts?
2. **Bootstrapper test** — Does it assemble relevant context correctly?
3. **Knowledge store test** — Does the index.json maintain consistency?
4. **Confidence calibration test** — Do high-confidence facts actually prove true in later sessions?
5. **Deduplication test** — Are duplicate facts across sessions handled?
6. **Staleness test** — How does the system handle outdated knowledge?
---
## Security Considerations
1. **No secrets in knowledge store** — The harvester must filter out API keys, tokens, and credentials from extracted facts. The prompt constraints mention this but there is no automated guard.
2. **Knowledge poisoning** — A malicious or corrupted session could inject false facts. Confidence scoring partially mitigates this, but there is no verification step.
3. **Access control** — The knowledge store has no access control. Any process that can read the directory can read all facts. In a multi-tenant setup, this is a concern.
4. **Transcript privacy** — Session transcripts may contain user data. The harvester must not extract personally identifiable information into the knowledge store.
---
## The 100x Path (from README)
```
Month 1: 15,000 facts, sessions 20% faster
Month 2: 45,000 facts, sessions 40% faster, first-try success up 30%
Month 3: 90,000 facts, fleet measurably smarter per token
```
Each new session is better than the last. The intelligence compounds.
---
*Generated by codebase-genome pipeline. Ref: timmy-home#676.*

317
scripts/dedup.py Normal file
View File

@@ -0,0 +1,317 @@
#!/usr/bin/env python3
"""
dedup.py — Knowledge deduplication: content hash + semantic similarity.
Deduplicates harvested knowledge entries to avoid training on duplicates.
Uses content hashing for exact matches and token overlap for near-duplicates.
Usage:
python3 dedup.py --input knowledge/index.json --output knowledge/index_deduped.json
python3 dedup.py --input knowledge/index.json --dry-run
python3 dedup.py --test # Run built-in dedup test
"""
import argparse
import hashlib
import json
import re
import sys
from pathlib import Path
from typing import List, Dict, Optional, Tuple
def normalize_text(text: str) -> str:
"""Normalize text for hashing: lowercase, collapse whitespace, strip."""
text = text.lower().strip()
text = re.sub(r'\s+', ' ', text)
return text
def content_hash(text: str) -> str:
"""SHA256 hash of normalized text for exact dedup."""
normalized = normalize_text(text)
return hashlib.sha256(normalized.encode('utf-8')).hexdigest()
def tokenize(text: str) -> set:
"""Simple tokenizer: lowercase words, 3+ chars."""
words = re.findall(r'[a-z0-9_]{3,}', text.lower())
return set(words)
def token_similarity(a: str, b: str) -> float:
"""Token-based Jaccard similarity (0.0-1.0).
Fast local alternative to embedding similarity.
Good enough for near-duplicate detection.
"""
tokens_a = tokenize(a)
tokens_b = tokenize(b)
if not tokens_a or not tokens_b:
return 0.0
intersection = tokens_a & tokens_b
union = tokens_a | tokens_b
return len(intersection) / len(union)
def quality_score(fact: dict) -> float:
"""Compute quality score for merge ranking.
Higher is better. Factors:
- confidence (0-1)
- source_count (more confirmations = better)
- has tags (richer metadata)
"""
confidence = fact.get('confidence', 0.5)
source_count = fact.get('source_count', 1)
has_tags = 1.0 if fact.get('tags') else 0.0
has_related = 1.0 if fact.get('related') else 0.0
# Weighted composite
score = (
confidence * 0.5 +
min(source_count / 10, 1.0) * 0.3 +
has_tags * 0.1 +
has_related * 0.1
)
return round(score, 4)
def merge_facts(keep: dict, drop: dict) -> dict:
"""Merge two near-duplicate facts, keeping higher-quality fields.
The 'keep' fact is enriched with metadata from 'drop'.
"""
# Merge tags (union)
keep_tags = set(keep.get('tags', []))
drop_tags = set(drop.get('tags', []))
keep['tags'] = sorted(keep_tags | drop_tags)
# Merge related (union)
keep_related = set(keep.get('related', []))
drop_related = set(drop.get('related', []))
keep['related'] = sorted(keep_related | drop_related)
# Update source_count (sum)
keep['source_count'] = keep.get('source_count', 1) + drop.get('source_count', 1)
# Update confidence (max — we've now seen it from multiple sources)
keep['confidence'] = max(keep.get('confidence', 0), drop.get('confidence', 0))
# Track that we merged
if '_merged_from' not in keep:
keep['_merged_from'] = []
keep['_merged_from'].append(drop.get('id', 'unknown'))
return keep
def dedup_facts(
facts: List[dict],
exact_threshold: float = 1.0,
near_threshold: float = 0.95,
dry_run: bool = False,
) -> Tuple[List[dict], dict]:
"""Deduplicate a list of knowledge facts.
Args:
facts: List of fact dicts (from index.json)
exact_threshold: Hash match = exact duplicate
near_threshold: Token similarity above this = near-duplicate
dry_run: If True, don't modify, just report
Returns:
(deduped_facts, stats_dict)
"""
if not facts:
return [], {"total": 0, "exact_dupes": 0, "near_dupes": 0, "unique": 0}
# Phase 1: Exact dedup by content hash
hash_seen = {} # hash -> index in deduped list
exact_dupes = 0
deduped = []
for fact in facts:
text = fact.get('fact', '')
h = content_hash(text)
if h in hash_seen:
# Exact duplicate — merge metadata into existing
existing_idx = hash_seen[h]
if not dry_run:
deduped[existing_idx] = merge_facts(deduped[existing_idx], fact)
exact_dupes += 1
else:
hash_seen[h] = len(deduped)
deduped.append(fact)
# Phase 2: Near-dup by token similarity
near_dupes = 0
i = 0
while i < len(deduped):
j = i + 1
while j < len(deduped):
sim = token_similarity(deduped[i].get('fact', ''), deduped[j].get('fact', ''))
if sim >= near_threshold:
# Near-duplicate — keep higher quality
q_i = quality_score(deduped[i])
q_j = quality_score(deduped[j])
if q_i >= q_j:
if not dry_run:
deduped[i] = merge_facts(deduped[i], deduped[j])
deduped.pop(j)
else:
# j is higher quality — merge i into j, then remove i
if not dry_run:
deduped[j] = merge_facts(deduped[j], deduped[i])
deduped.pop(i)
break # i changed, restart inner loop
near_dupes += 1
else:
j += 1
i += 1
stats = {
"total": len(facts),
"exact_dupes": exact_dupes,
"near_dupes": near_dupes,
"unique": len(deduped),
"removed": len(facts) - len(deduped),
}
return deduped, stats
def dedup_index_file(
input_path: str,
output_path: Optional[str] = None,
near_threshold: float = 0.95,
dry_run: bool = False,
) -> dict:
"""Deduplicate an index.json file.
Args:
input_path: Path to index.json
output_path: Where to write deduped file (default: overwrite input)
near_threshold: Token similarity threshold for near-dupes
dry_run: Report only, don't write
Returns stats dict.
"""
path = Path(input_path)
if not path.exists():
raise FileNotFoundError(f"Index file not found: {input_path}")
with open(path) as f:
data = json.load(f)
facts = data.get('facts', [])
deduped, stats = dedup_facts(facts, near_threshold=near_threshold, dry_run=dry_run)
if not dry_run:
data['facts'] = deduped
data['total_facts'] = len(deduped)
data['last_dedup'] = __import__('datetime').datetime.now(
__import__('datetime').timezone.utc
).isoformat()
out_path = Path(output_path) if output_path else path
with open(out_path, 'w') as f:
json.dump(data, f, indent=2, ensure_ascii=False)
return stats
def generate_test_duplicates(n: int = 20) -> List[dict]:
"""Generate test facts with intentional duplicates for testing.
Creates n unique facts plus n/4 exact dupes and n/4 near-dupes.
"""
import random
random.seed(42)
unique_facts = []
for i in range(n):
topic = random.choice(["git", "python", "docker", "rust", "nginx"])
tip = random.choice(["use verbose flags", "check logs first", "restart service", "clear cache", "update config"])
unique_facts.append({
"id": f"test:fact:{i:03d}",
"fact": f"When working with {topic}, always {tip} before deploying.",
"category": "fact",
"domain": "test",
"confidence": round(random.uniform(0.5, 1.0), 2),
"source_count": random.randint(1, 5),
"tags": [topic, "test"],
})
# Add exact duplicates (same text, different IDs)
duped = list(unique_facts)
for i in range(n // 4):
original = unique_facts[i]
dupe = dict(original)
dupe["id"] = f"test:fact:dup{i:03d}"
dupe["confidence"] = round(random.uniform(0.3, 0.8), 2)
duped.append(dupe)
# Add near-duplicates (slightly different phrasing)
for i in range(n // 4):
original = unique_facts[i]
near = dict(original)
near["id"] = f"test:fact:near{i:03d}"
near["fact"] = original["fact"].replace("always", "should").replace("before deploying", "prior to deployment")
near["confidence"] = round(random.uniform(0.4, 0.9), 2)
duped.append(near)
return duped
def main():
parser = argparse.ArgumentParser(description="Knowledge deduplication")
parser.add_argument("--input", help="Path to index.json")
parser.add_argument("--output", help="Output path (default: overwrite input)")
parser.add_argument("--threshold", type=float, default=0.95,
help="Near-dup similarity threshold (default: 0.95)")
parser.add_argument("--dry-run", action="store_true", help="Report only, don't write")
parser.add_argument("--test", action="store_true", help="Run built-in dedup test")
parser.add_argument("--json", action="store_true", help="JSON output")
args = parser.parse_args()
if args.test:
test_facts = generate_test_duplicates(20)
print(f"Generated {len(test_facts)} test facts (20 unique + dupes)")
deduped, stats = dedup_facts(test_facts, near_threshold=args.threshold)
print(f"\nDedup results:")
print(f" Total input: {stats['total']}")
print(f" Exact dupes: {stats['exact_dupes']}")
print(f" Near dupes: {stats['near_dupes']}")
print(f" Unique output: {stats['unique']}")
print(f" Removed: {stats['removed']}")
# Verify: should have ~20 unique (some merged)
assert stats['unique'] <= 20, f"Too many unique: {stats['unique']} > 20"
assert stats['unique'] >= 15, f"Too few unique: {stats['unique']} < 15"
assert stats['removed'] > 0, "No duplicates removed"
print("\nOK: Dedup test passed")
return
if not args.input:
print("ERROR: Provide --input or --test")
sys.exit(1)
stats = dedup_index_file(args.input, args.output, args.threshold, args.dry_run)
if args.json:
print(json.dumps(stats, indent=2))
else:
print(f"Dedup results:")
print(f" Total input: {stats['total']}")
print(f" Exact dupes: {stats['exact_dupes']}")
print(f" Near dupes: {stats['near_dupes']}")
print(f" Unique output: {stats['unique']}")
print(f" Removed: {stats['removed']}")
if args.dry_run:
print(" (dry run — no changes written)")
if __name__ == "__main__":
main()

207
tests/test_dedup.py Normal file
View File

@@ -0,0 +1,207 @@
"""Tests for knowledge deduplication module (Issue #196)."""
import json
import sys
from pathlib import Path
import pytest
sys.path.insert(0, str(Path(__file__).parent.parent / "scripts"))
from dedup import (
normalize_text,
content_hash,
tokenize,
token_similarity,
quality_score,
merge_facts,
dedup_facts,
generate_test_duplicates,
)
class TestNormalize:
def test_lowercases(self):
assert normalize_text("Hello World") == "hello world"
def test_collapses_whitespace(self):
assert normalize_text(" hello world ") == "hello world"
def test_strips(self):
assert normalize_text(" text ") == "text"
class TestContentHash:
def test_deterministic(self):
h1 = content_hash("Hello World")
h2 = content_hash("hello world")
h3 = content_hash(" Hello World ")
assert h1 == h2 == h3
def test_different_texts(self):
h1 = content_hash("Hello")
h2 = content_hash("World")
assert h1 != h2
def test_returns_hex(self):
h = content_hash("test")
assert len(h) == 64 # SHA256
assert all(c in '0123456789abcdef' for c in h)
class TestTokenize:
def test_extracts_words(self):
tokens = tokenize("Hello World Test")
assert "hello" in tokens
assert "world" in tokens
assert "test" in tokens
def test_skips_short_words(self):
tokens = tokenize("a to is the hello")
assert "a" not in tokens
assert "to" not in tokens
assert "hello" in tokens
def test_returns_set(self):
tokens = tokenize("hello hello world")
assert isinstance(tokens, set)
assert len(tokens) == 2
class TestTokenSimilarity:
def test_identical(self):
assert token_similarity("hello world", "hello world") == 1.0
def test_no_overlap(self):
assert token_similarity("alpha beta", "gamma delta") == 0.0
def test_partial_overlap(self):
sim = token_similarity("hello world test", "hello universe test")
assert 0.3 < sim < 0.7
def test_empty(self):
assert token_similarity("", "hello") == 0.0
assert token_similarity("hello", "") == 0.0
def test_symmetric(self):
a = "hello world test"
b = "hello universe test"
assert token_similarity(a, b) == token_similarity(b, a)
class TestQualityScore:
def test_high_confidence(self):
fact = {"confidence": 0.95, "source_count": 5, "tags": ["test"], "related": ["x"]}
score = quality_score(fact)
assert score > 0.7
def test_low_confidence(self):
fact = {"confidence": 0.3, "source_count": 1}
score = quality_score(fact)
assert score < 0.5
def test_defaults(self):
score = quality_score({})
assert 0 < score < 1
class TestMergeFacts:
def test_merges_tags(self):
keep = {"id": "a", "fact": "test", "tags": ["git"], "confidence": 0.9}
drop = {"id": "b", "fact": "test", "tags": ["python"], "confidence": 0.8}
merged = merge_facts(keep, drop)
assert "git" in merged["tags"]
assert "python" in merged["tags"]
def test_merges_source_count(self):
keep = {"id": "a", "fact": "test", "source_count": 3}
drop = {"id": "b", "fact": "test", "source_count": 2}
merged = merge_facts(keep, drop)
assert merged["source_count"] == 5
def test_keeps_higher_confidence(self):
keep = {"id": "a", "fact": "test", "confidence": 0.7}
drop = {"id": "b", "fact": "test", "confidence": 0.9}
merged = merge_facts(keep, drop)
assert merged["confidence"] == 0.9
def test_tracks_merged_from(self):
keep = {"id": "a", "fact": "test"}
drop = {"id": "b", "fact": "test"}
merged = merge_facts(keep, drop)
assert "b" in merged["_merged_from"]
class TestDedupFacts:
def test_removes_exact_dupes(self):
facts = [
{"id": "1", "fact": "Always use git rebase"},
{"id": "2", "fact": "Always use git rebase"}, # exact dupe
{"id": "3", "fact": "Check logs first"},
]
deduped, stats = dedup_facts(facts)
assert stats["exact_dupes"] == 1
assert stats["unique"] == 2
def test_removes_near_dupes(self):
facts = [
{"id": "1", "fact": "Always check logs before deploying to production server"},
{"id": "2", "fact": "Always check logs before deploying to production environment"},
{"id": "3", "fact": "Use docker compose for local development environments"},
]
deduped, stats = dedup_facts(facts, near_threshold=0.5)
assert stats["near_dupes"] >= 1
assert stats["unique"] == 2
def test_preserves_unique(self):
facts = [
{"id": "1", "fact": "Use git rebase for clean history"},
{"id": "2", "fact": "Docker containers should be stateless"},
{"id": "3", "fact": "Always write tests before code"},
]
deduped, stats = dedup_facts(facts)
assert stats["unique"] == 3
assert stats["removed"] == 0
def test_empty_input(self):
deduped, stats = dedup_facts([])
assert stats["total"] == 0
assert stats["unique"] == 0
def test_keeps_higher_quality_near_dup(self):
facts = [
{"id": "1", "fact": "Check logs before deploying to production server", "confidence": 0.5, "source_count": 1},
{"id": "2", "fact": "Check logs before deploying to production environment", "confidence": 0.9, "source_count": 5, "tags": ["ops"]},
]
deduped, stats = dedup_facts(facts, near_threshold=0.5)
assert stats["unique"] == 1
# Higher quality fact should be kept
assert deduped[0]["confidence"] == 0.9
def test_dry_run_does_not_modify(self):
facts = [
{"id": "1", "fact": "Same text"},
{"id": "2", "fact": "Same text"},
]
deduped, stats = dedup_facts(facts, dry_run=True)
assert stats["exact_dupes"] == 1
# In dry_run, merge_facts is skipped so facts aren't modified
assert len(deduped) == 1
class TestGenerateTestDuplicates:
def test_generates_correct_count(self):
facts = generate_test_duplicates(20)
assert len(facts) > 20 # 20 unique + duplicates
def test_has_exact_dupes(self):
facts = generate_test_duplicates(20)
hashes = [content_hash(f["fact"]) for f in facts]
# Should have some duplicate hashes
assert len(hashes) != len(set(hashes))
def test_dedup_removes_dupes(self):
facts = generate_test_duplicates(20)
deduped, stats = dedup_facts(facts)
assert stats["unique"] <= 20
assert stats["removed"] > 0