Core Modules

Knowledge Base

The Knowledge Base module powers Retrieval-Augmented Generation (RAG) for your AI assistants. Content is chunked, embedded, and stored in Qdrant Cloud for fast semantic search.

How it works

Ingest — upload documents, URLs, GitHub repos, or plain text.
Chunk & Embed — the backend splits content into chunks and generates vector embeddings using your configured AI provider (OpenAI, Anthropic, Google, or local).
Store — embeddings are stored in Qdrant Cloud; metadata is stored in Firestore.
Retrieve — at chat time, the most relevant chunks are retrieved and injected into the AI context window.

Ingesting content

From URLs

Crawl and ingest web pages. Supports multiple URLs in a single request:

Shell

curl -X POST https://api.cognitiondesk.com/knowledge-base/ingest/url \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
    "urls": ["https://example.com/docs", "https://example.com/faq"],
    "metadata": { "source": "website", "category": "support" }
  }'

From GitHub

Ingest an entire repository. You can filter by file patterns:

Shell

curl -X POST https://api.cognitiondesk.com/knowledge-base/ingest/github \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
    "repoUrl": "https://github.com/your-org/your-repo",
    "branch": "main",
    "filePatterns": ["**/*.md", "**/*.txt"]
  }'

Document upload

First upload the document using POST /upload, then ingest the returned file ID:

1. POST /upload with multipart form data → returns fileId
2. POST /knowledge-base/ingest/document with { fileId }

Semantic search

Run a standalone semantic search against your knowledge base — useful for building custom retrieval pipelines:

Shell

curl -X POST https://api.cognitiondesk.com/knowledge-base/search \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "How do I reset my password?",
    "limit": 5,
    "threshold": 0.75
  }'

Response

JSON

{
  "results": [
    {
      "id": "chunk_abc123",
      "content": "To reset your password, navigate to Settings > Security...",
      "score": 0.92,
      "metadata": { "source": "website", "url": "https://example.com/faq" }
    }
  ],
  "query": "How do I reset my password?",
  "total": 5
}

Endpoints

Method	Endpoint	Auth	Description
GET	`/knowledge-base/stats`	Firebase Token	Total entries, chunks, and storage usage
GET	`/knowledge-base/entries`	Firebase Token	List entries (filter by source, date, category)
GET	`/knowledge-base/entries/:id`	Firebase Token	Get a specific entry by ID
PATCH	`/knowledge-base/entries/:id`	Firebase Token	Update entry metadata
DELETE	`/knowledge-base/entries/:id`	Firebase Token	Delete entry from Firestore + Qdrant
POST	`/knowledge-base/ingest/url`	Firebase Token	Ingest content from one or more URLs
POST	`/knowledge-base/ingest/github`	Firebase Token	Ingest entire GitHub repositories
POST	`/knowledge-base/ingest/text`	Firebase Token	Ingest plain text content
POST	`/knowledge-base/ingest/image`	Firebase Token	Ingest an image (vision models)
POST	`/knowledge-base/ingest/document`	Firebase Token	Ingest an uploaded document (PDF, DOCX)
GET	`/knowledge-base/chunks`	Firebase Token	Explore raw vector chunks in Qdrant
POST	`/knowledge-base/search`	Firebase Token	Semantic similarity search