firefrost-operations-manual/docs/tasks/task-093-trinity-codex/references/skill-seekers-qdrant.md

# Skill Seekers + Qdrant Integration

**Source:** https://github.com/yusufkaraaslan/Skill_Seekers  
**License:** MIT  
**Gitea Fork:** https://git.firefrostgaming.com/firefrost-gaming/skill-seekers-reference

## Overview

Skill Seekers converts documentation sites, GitHub repos, PDFs, and 17+ source types into structured knowledge assets ready for RAG pipelines. This is directly applicable to Trinity Codex.

## Installation

```bash
pip install skill-seekers
```

## Quick Start

```bash
# Convert docs to skill
skill-seekers create https://docs.example.com/

# Package for Qdrant
skill-seekers package output/example --target qdrant
```

## Supported Sources (17 types)

- Documentation websites
- GitHub repositories
- PDF documents
- Word documents (.docx)
- EPUB e-books
- Jupyter Notebooks
- OpenAPI specs
- PowerPoint presentations
- AsciiDoc documents
- HTML files
- RSS/Atom feeds
- Man pages
- YouTube videos (with `skill-seekers[video]`)

## Qdrant Pipeline

### Step 1: Generate Skill

```python
#!/usr/bin/env python3
import subprocess
from pathlib import Path

# Scrape documentation
subprocess.run([
    "skill-seekers", "scrape",
    "--config", "configs/your-config.json",
    "--max-pages", "20"
], check=True)

# Package for Qdrant
subprocess.run([
    "skill-seekers", "package",
    "output/your-skill",
    "--target", "qdrant"
], check=True)

output = Path("output/your-skill-qdrant.json")
print(f"Ready: {output} ({output.stat().st_size/1024:.1f} KB)")
```

### Step 2: Upload to Qdrant

```python
#!/usr/bin/env python3
import json
from qdrant_client import QdrantClient
from qdrant_client.models import Distance, VectorParams, PointStruct

# Connect to Qdrant (our instance will be on TX1)
client = QdrantClient(url="http://localhost:6333")

# Load packaged data
with open("output/your-skill-qdrant.json") as f:
    data = json.load(f)

collection_name = data["collection_name"]
config = data["config"]

# Create collection
client.create_collection(
    collection_name=collection_name,
    vectors_config=VectorParams(
        size=config["vector_size"],
        distance=Distance.COSINE
    )
)

# Upload points (add real embeddings in production)
points = []
for point in data["points"]:
    points.append(PointStruct(
        id=point["id"],
        vector=[0.0] * config["vector_size"],  # Replace with real embeddings
        payload=point["payload"]
    ))

client.upsert(collection_name=collection_name, points=points)
print(f"Uploaded {len(points)} points to {collection_name}")
```

### Step 3: Query

```python
#!/usr/bin/env python3
from qdrant_client import QdrantClient
from qdrant_client.models import Filter, FieldCondition, MatchValue

client = QdrantClient(url="http://localhost:6333")
collection_name = "your-collection"

# Filter by category
result = client.scroll(
    collection_name=collection_name,
    scroll_filter=Filter(
        must=[
            FieldCondition(
                key="category",
                match=MatchValue(value="api")
            )
        ]
    ),
    limit=5
)

for point in result[0]:
    print(f"- {point.payload['file']}: {point.payload['content'][:100]}...")
```

## Trinity Codex Application

### Phase 1: Documentation Ingestion

Convert key Firefrost documentation sources:

```bash
# Pterodactyl docs
skill-seekers create https://pterodactyl.io/project/introduction.html
skill-seekers package output/pterodactyl --target qdrant

# Minecraft Wiki (modding)
skill-seekers create https://minecraft.wiki/w/Mods

# Operations Manual (local)
skill-seekers create ./docs/
skill-seekers package output/docs --target qdrant
```

### Phase 2: Vector Database Setup

Qdrant runs on TX1 (38.68.14.26) alongside Dify:

```bash
# Docker deployment
docker run -d \
  --name qdrant \
  -p 6333:6333 \
  -v /opt/qdrant/storage:/qdrant/storage \
  qdrant/qdrant:latest
```

### Phase 3: Dify Integration

Dify connects to Qdrant for RAG queries. See Dify documentation for knowledge base configuration.

## Key Features for Firefrost

| Feature | Benefit |
|---------|---------|
| Multi-source ingestion | Combine wiki, docs, PDFs into one knowledge base |
| Qdrant-native output | Direct integration with our planned stack |
| Smart chunking | Preserves code blocks and context |
| Metadata preservation | Category, file, type fields for filtering |
| 500+ line SKILL.md | High-quality Claude skills from any source |

## Resources

- **Full repo:** https://git.firefrostgaming.com/firefrost-gaming/skill-seekers-reference
- **Original:** https://github.com/yusufkaraaslan/Skill_Seekers
- **Website:** https://skillseekersweb.com/
- **Qdrant Docs:** https://qdrant.tech/documentation/

---

*Added by Chronicler #73 on 2026-04-09*