---
license: apache-2.0
base_model: edgeimpulse/edgeimpulse-docs-qwen-0.5b
pipeline_tag: text-generation
library_name: transformers
tags:
  - edge-impulse
  - rag
  - retrieval-augmented-generation
  - faiss
  - qwen
  - api
  - documentation
  - tinyml
  - edge-ai
---

# Edge Impulse API Docs — RAG Assistant

A focused retrieval-augmented assistant for the **Edge Impulse API reference**
(Studio API, ingestion API, remote-management API). Same runtime as the full
[`edgeimpulse/edgeimpulse-docs-rag`](https://huggingface.co/edgeimpulse/edgeimpulse-docs-rag)
assistant, but the knowledge base is restricted to the API documentation so
retrieval stays tightly on-topic for integration and automation questions.

- **Scope:** Edge Impulse API reference only (`apis`, `apis_studio`,
  `apis_ingestion`, `apis_remote-management`).
- **Retrieval:** FAISS (inner-product) over `data/index`, embedded with
  `sentence-transformers/all-MiniLM-L6-v2` (384-dim).
- **Generation:** [`edgeimpulse/edgeimpulse-docs-qwen-0.5b`](https://huggingface.co/edgeimpulse/edgeimpulse-docs-qwen-0.5b)
  via any OpenAI-compatible endpoint (llama.cpp or Ollama).

This repo ships only the prebuilt (API-scoped) index and the inference code — no
raw corpus and no index-building pipeline.

## Contents

| File | Purpose |
| --- | --- |
| `data/index/edge_impulse_docs.faiss` | FAISS index of the API-reference chunks |
| `data/index/chunks.pkl` | Chunk text + source metadata (aligned to the index) |
| `data/index/metadata.json` | Embedding model + index parameters (`scope: edge-impulse-api-reference`) |
| `rag.py` | Retrieval + grounded generation (CLI + importable) |
| `serve.py` | Minimal Flask HTTP API (`POST /ask`) |
| `requirements.txt` | Runtime dependencies |

## Quickstart

```bash
pip install -r requirements.txt
hf download edgeimpulse/edgeimpulse-api-docs-rag --local-dir edgeimpulse-api-docs-rag
cd edgeimpulse-api-docs-rag

# start the generator (llama.cpp)
hf download edgeimpulse/edgeimpulse-docs-qwen-0.5b qwen-edgeai-q4_k_m.gguf --local-dir .
llama-server -m qwen-edgeai-q4_k_m.gguf -c 4096 --port 8080 --jinja

# ask an API question
python rag.py "How do I create an Edge Impulse API key?"
python rag.py "How do I upload data with the ingestion API?" --no-generate
```

Serve over HTTP:

```bash
python serve.py --host 0.0.0.0 --port 8000
curl -s localhost:8000/ask -H 'content-type: application/json' \
  -d '{"question": "How do I classify an image via the API?"}'
```

## Configuration

`rag.py` honours `RAG_INDEX_DIR`, `RAG_API_BASE`, and `RAG_MODEL` (see the full
assistant's card for details). For Ollama:

```bash
export RAG_API_BASE=http://127.0.0.1:11434/v1
export RAG_MODEL=hf.co/edgeimpulse/edgeimpulse-docs-qwen-0.5b
```

## Notes

Because the corpus is intentionally narrow, questions outside the API reference
may return "not in context" — that is by design. For general docs questions
(devices, deployment targets, ML concepts) use the full
[`edgeimpulse/edgeimpulse-docs-rag`](https://huggingface.co/edgeimpulse/edgeimpulse-docs-rag)
assistant instead.

## License

Apache-2.0. Documentation content belongs to Edge Impulse.