Public Embedding & Reranking API endpoints powered by Lumox (Strix Halo)
| Filename | Latest commit message | Latest commit date |
|---|---|---|
#2: Document rate limits and batch sizes (embedding.md, reranking.md) #3: Mark embed_api_* fields as custom build, not official QMD (openclaw.md) #4: Add MRL explanation with dimension table and examples (embedding.md) #5: Add dimension mismatch warning for fallback models (openclaw.md) |
||
| docs | ||
| examples | ||
| README.md | ||
Lumox Retrieval API
Embedding & Reranking endpoints powered by Lumox (AMD Strix Halo, Vulkan backend).
🇩🇪 Deutsch | 🇬🇧 English
📖 Dokumentation
| Endpoint | URL | Modell | Dimensionen | Max. Kontext |
|---|---|---|---|---|
| Embedding | https://embedding.nextmock.de/v1/embeddings |
Qwen3-Embedding-4B | 2560 | 32.768 Token |
| Reranking | https://reranking.nextmock.de/v1/rerank |
Qwen3-Reranker-4B | — | 8.192 Token |
API-Key
Auf Anfrage. Embedding und Reranking nutzen denselben Key.
Quick Start
# Embedding
curl -X POST https://embedding.nextmock.de/v1/embeddings \
-H "Authorization: Bearer <API_KEY>" \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen3-Embedding-4B",
"input": "Filtertechnik in der industriellen Wasseraufbereitung"
}'
# Reranking
curl -X POST https://reranking.nextmock.de/v1/rerank \
-H "Authorization: Bearer <API_KEY>" \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen3-Reranker-4B",
"query": "Was ist Reverse Osmose?",
"documents": [
"Reverse Osmose ist ein Filtrationsverfahren mit Semipermeable Membran",
"Der BMW M3 hat 480 PS",
"Aktivkohlefilter entfernen organische Verbindungen aus Wasser"
],
"top_n": 2
}'
Integration
- OpenClaw / QMD – Integration via QMD Memory Backend
- Python – OpenAI SDK & Requests
- PostgreSQL + pgvector – Eigene Retrieval-Pipeline bauen
- Embedding Details – Parameter, Response-Format
- Reranking Details – Parameter, Response-Format
- Beispiele – Copy-paste-fertige Scripts
🇬🇧 English | 🇩🇪 Deutsch
📖 Documentation
| Endpoint | URL | Model | Dimensions | Max. Context |
|---|---|---|---|---|
| Embedding | https://embedding.nextmock.de/v1/embeddings |
Qwen3-Embedding-4B | 2560 | 32,768 tokens |
| Reranking | https://reranking.nextmock.de/v1/rerank |
Qwen3-Reranker-4B | — | 8,192 tokens |
API Key
On request. Embedding and Reranking share the same key.
Quick Start
# Embedding
curl -X POST https://embedding.nextmock.de/v1/embeddings \
-H "Authorization: Bearer <API_KEY>" \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen3-Embedding-4B",
"input": "Industrial water treatment filtration technology"
}'
# Reranking
curl -X POST https://reranking.nextmock.de/v1/rerank \
-H "Authorization: Bearer <API_KEY>" \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen3-Reranker-4B",
"query": "What is reverse osmosis?",
"documents": [
"Reverse osmosis is a filtration process using a semipermeable membrane",
"The BMW M3 has 480 horsepower",
"Activated carbon filters remove organic compounds from water"
],
"top_n": 2
}'
Integration
- OpenClaw / QMD – Integration via QMD memory backend
- Python – OpenAI SDK & Requests
- PostgreSQL + pgvector – Build your own retrieval pipeline
- Embedding Details – Parameters, response format
- Reranking Details – Parameters, response format
- Examples – Ready-to-use scripts
Specs
| Property | Value |
|---|---|
| Backend | Vulkan (AMD Radeon 8060S / Strix Halo) |
| API Format | OpenAI-compatible |
| Protocol | HTTPS |
| Auth | Bearer token |
Contact
| Role | Name | Contact |
|---|---|---|
| Operator & Infrastructure | Chris | — |
| Agent & Documentation | Shadow | — |
Powered by Lumox — AMD Strix Halo GPU server.