Local inference
Local model inference as a platform service: pinned llama.cpp engines on the platform's own hardware.
partAI9 operationsid llama
How it sits
A part does one thing, and any other part can call it. The counts and the links below are from the platform's spec and the other services' own descriptions.
| Operations | 9 GET 6 · POST 3 |
| Resources | 7 |
| Leaned on by | no other service names it |
| Leans on | none named |
| In recipes | none yet |
What it can do
9 operations in 7 resources, as the spec declares them. Each row is one call; paths are relative to the part.
chat
| Method | Path | Does |
|---|---|---|
| POST | /api/v1/chat | Chat (stream or detach for long work) |
| POST | /v1/chat/completions | OpenAI-compatible chat |
2 operations
health
| Method | Path | Does |
|---|---|---|
| GET | /health | Basic health check |
| GET | /health/detailed | Detailed health check |
2 operations
decide
| Method | Path | Does |
|---|---|---|
| POST | /api/v1/decide | Typed decisions (decision models) |
1 operation
jobs
| Method | Path | Does |
|---|---|---|
| GET | /api/v1/jobs/{job_id} | Detached job state and result |
1 operation
models
| Method | Path | Does |
|---|---|---|
| GET | /api/v1/models | Models and their live state |
1 operation
prime
| Method | Path | Does |
|---|---|---|
| GET | /prime | Prime application caches |
1 operation
status
| Method | Path | Does |
|---|---|---|
| GET | /status | Service status overview |
1 operation
All the partsAIThe platform's own id for this service is cbllama.
forge-catalog.json, generated Oct 30, 2026