local-llm-worker
by JoblessJoe
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| LLW_API | No | API type: auto, ollama, or openai. Auto probes /api/version. | auto |
| LLW_MODEL | No | Default model for both tools. | |
| LLW_API_KEY | No | Sent as Authorization: Bearer. Shown as (set), never printed. Plugin users can set it under the plugin's settings instead, which keeps it in the OS keychain. | |
| LLW_HEADERS | No | Extra HTTP headers for a gateway, e.g. {"X-Api-Key": "..."}. Values shown as (set). | {} |
| LLW_NUM_CTX | No | Context window (Ollama). | 32768 |
| LLW_AUTO_USE | No | Tools Claude uses on its own, via a short session-start reminder. delegate is opt-in. Every tool still works when you ask for it; [] turns the reminder off. | offload, research |
| LLW_BASE_URL | No | Backend root. A trailing `/v1` is stripped. | http://localhost:11434 |
| LLW_LOG_PATH | No | Run log; "" disables. | ~/.local-llm-worker/runs.jsonl |
| LLW_LINK_DIRS | No | Symlinked into each worktree. | node_modules, .venv, venv, vendor |
| LLW_KEEP_ALIVE | No | Ollama only: how long the model stays loaded, e.g. "30m" or -1 (forever). "" keeps the server default (5 min). | |
| LLW_MAX_TOKENS | No | Output limit per answer, sent on both APIs (num_predict on Ollama). A cut-off answer is reported, not tested. | 8192 |
| LLW_SEARCH_URL | No | SearXNG instance for research (JSON output enabled). | |
| LLW_TIMEOUT_MS | No | Per LLM request, counted from slot acquisition. | 600000 |
| LLW_CONCURRENCY | No | Max in-flight LLM requests per server process. | 4 |
| LLW_MAX_ATTEMPTS | No | Delegate rounds (cap 10). | 3 |
| LLW_OFFLOAD_MODEL | No | Override for offload (e.g. long-context). | |
| LLW_DELEGATE_MODEL | No | Override for delegate (e.g. a coder). | |
| LLW_RESEARCH_MODEL | No | Override for research (e.g. long-context). | |
| LLW_PAGE_TIMEOUT_MS | No | Per page download and per search. | 30000 |
| LLW_TEST_TIMEOUT_MS | No | Per test / command run. | 600000 |
| LLW_RESEARCH_SOURCES | No | Pages research reads per search (max 10). | 3 |
| LLW_ALLOW_PRIVATE_URLS | No | Let research read pages on loopback, link-local or private addresses (the search_url itself is always allowed). | false |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Server capabilities have not been inspected yet.
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
This server cannot be deployed
Maintenance
ActivityNo data
ResponsivenessNo issues