mirror of
https://github.com/Kaelio/ktx.git
synced 2026-06-13 08:15:14 +02:00
62 lines
2.1 KiB
Text
62 lines
2.1 KiB
Text
|
---
|
||
|
|
title: LLM configuration
|
||
|
|
description: Configure KTX LLM providers, model roles, and prompt caching.
|
||
|
|
---
|
||
|
|
|
||
|
|
KTX uses the top-level `llm` block in `ktx.yaml` for text generation,
|
||
|
|
structured extraction, and ingest or memory agent loops.
|
||
|
|
|
||
|
|
## Backends
|
||
|
|
|
||
|
|
Set `llm.provider.backend` to one of these values:
|
||
|
|
|
||
|
|
- `anthropic`: Use the Anthropic API through `ANTHROPIC_API_KEY` or the
|
||
|
|
configured `api_key` reference.
|
||
|
|
- `vertex`: Use Vertex AI Anthropic models through Google Cloud credentials.
|
||
|
|
- `gateway`: Use AI Gateway-compatible Anthropic model ids.
|
||
|
|
- `claude-code`: Use your local Claude Code session through the Claude Agent
|
||
|
|
SDK. KTX removes provider-routing environment variables from Claude Code
|
||
|
|
child processes, so this backend doesn't silently fall back to
|
||
|
|
`ANTHROPIC_API_KEY`, Vertex, Gateway, or Bedrock credentials.
|
||
|
|
|
||
|
|
## Claude Code
|
||
|
|
|
||
|
|
Use aliases or full Claude model IDs in `llm.models`:
|
||
|
|
|
||
|
|
```yaml
|
||
|
|
llm:
|
||
|
|
provider:
|
||
|
|
backend: claude-code
|
||
|
|
models:
|
||
|
|
default: sonnet
|
||
|
|
triage: haiku
|
||
|
|
candidateExtraction: sonnet
|
||
|
|
curator: sonnet
|
||
|
|
reconcile: sonnet
|
||
|
|
repair: sonnet
|
||
|
|
```
|
||
|
|
|
||
|
|
During setup, choose the Claude Code backend interactively or pass the model in
|
||
|
|
automation:
|
||
|
|
|
||
|
|
```bash
|
||
|
|
ktx setup --llm-backend claude-code --llm-model opus --no-input
|
||
|
|
```
|
||
|
|
|
||
|
|
For Claude Code, `sonnet`, `opus`, and `haiku` map to the current KTX defaults.
|
||
|
|
You can also pass a full Claude model ID, such as `claude-opus-4-7`.
|
||
|
|
|
||
|
|
`claude-code` keeps KTX tool boundaries intact. KTX exposes only the MCP tools
|
||
|
|
needed for the current KTX agent loop, disables Claude Code built-in tools,
|
||
|
|
keeps plugins empty, and denies every non-KTX tool request through
|
||
|
|
`canUseTool`. The Claude Agent SDK may still report host-discovered slash
|
||
|
|
commands, skills, and subagent names in init metadata; that metadata is not an
|
||
|
|
execution grant for KTX agent loops.
|
||
|
|
|
||
|
|
## Prompt caching
|
||
|
|
|
||
|
|
`llm.promptCaching` has partial parity on `claude-code`. KTX doesn't pass
|
||
|
|
Anthropic cache-control markers to the Claude Agent SDK. Status and doctor warn
|
||
|
|
when you configure prompt-cache TTL, tool, or history fields that the Claude
|
||
|
|
Agent SDK backend ignores.
|