@antseed/provider-local-llm
Provide local LLM capacity on the Antseed P2P network. Works with Ollama, llama.cpp, and any OpenAI-compatible local server.
Installation
The CLI installs the plugin when you add a provider that uses it:
antseed config seller add-provider local-llm --plugin local-llm
Usage
# Configure once
antseed config seller add-service local-llm llama3.2:3b \
--input 0 --output 0 \
--categories chat,fast,free
antseed seller settings set --max-buyers 1
# With Ollama (default)
antseed seller start
# With a custom endpoint
export LOCAL_LLM_BASE_URL=http://localhost:8080
antseed seller start
Configuration
| Key | Type | Required | Default | Description |
|---|---|---|---|---|
LOCAL_LLM_BASE_URL | string | No | http://localhost:11434 | Local LLM server URL |
LOCAL_LLM_API_KEY | secret | No | -- | Optional API key for local server |
ANTSEED_INPUT_USD_PER_MILLION | number | No | 0 | Input token price (USD per 1M) |
ANTSEED_OUTPUT_USD_PER_MILLION | number | No | 0 | Output token price (USD per 1M) |
ANTSEED_SERVICE_PRICING_JSON | string | No | -- | Per-service pricing as JSON |
ANTSEED_MAX_CONCURRENCY | number | No | 1 | Max concurrent requests |
ANTSEED_ALLOWED_SERVICES | string[] | No | -- | Comma-separated service allowlist |
How It Works
Relays requests to a local LLM server. Pricing defaults to 0 (free) since you're running your own hardware. The CLI sets the plugin's concurrency from seller.maxConcurrentBuyers, which defaults to 50, so lower it with antseed seller settings set --max-buyers 1 (or another number your hardware can handle) to avoid overloading local inference. The plugin's own default of 1 applies only when you load it outside the CLI.