f654c59dd5
- New variable honcho_deriver_autostart controls service state/enabled - honcho_deriver_enabled still controls whether Quadlet is created - Default autostart=false after discovering deriver burns tokens autonomously - Service is created but stopped/disabled; can be started manually if needed
160 lines
7.7 KiB
YAML
160 lines
7.7 KiB
YAML
---
|
|
# ============================================================================
|
|
# honcho role defaults
|
|
# ============================================================================
|
|
# Deploys Honcho (https://honcho.dev — plastic-labs/honcho) as a rootful
|
|
# Podman + Quadlet service on a single VM. Three containers:
|
|
#
|
|
# honcho-postgres pgvector-enabled PostgreSQL backing store
|
|
# honcho-api FastAPI server on :8000 (theory-of-mind read/write)
|
|
# honcho-deriver background worker that consumes the queue and calls
|
|
# out to the configured LLM provider for derivations
|
|
#
|
|
# All three share a user-defined podman network. State persists to named
|
|
# volumes. Traefik on lightning-lane terminates TLS and routes the
|
|
# `hall-of-presidents.local.mk-labs.cloud` host to honcho-api:8000.
|
|
#
|
|
# LLM provider for the first deployment is Anthropic Claude. Switching
|
|
# providers is a single env-var change (see DERIVER_MODEL_CONFIG__TRANSPORT
|
|
# / SUMMARY_MODEL_CONFIG__TRANSPORT and the LLM_*_API_KEY variables).
|
|
# ============================================================================
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Image pinning
|
|
# ---------------------------------------------------------------------------
|
|
# Honcho does not yet publish a Docker Hub image we trust; we use the
|
|
# GitHub Container Registry build. Pin to a digest-stable tag.
|
|
honcho_image: "ghcr.io/plastic-labs/honcho:latest"
|
|
# pgvector/pgvector image follows the upstream postgres version channel.
|
|
honcho_postgres_image: "docker.io/pgvector/pgvector:pg16"
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Container & network identifiers
|
|
# ---------------------------------------------------------------------------
|
|
honcho_container_name: honcho-api
|
|
honcho_deriver_container_name: honcho-deriver
|
|
honcho_postgres_container_name: honcho-postgres
|
|
honcho_network_name: honcho-net
|
|
|
|
# Named podman volumes
|
|
honcho_postgres_volume: honcho_postgres_data
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Networking
|
|
# ---------------------------------------------------------------------------
|
|
# Honcho's FastAPI default port is 8000. Bind to 0.0.0.0 so Traefik on
|
|
# lightning-lane can reach it; auth is enabled (JWT) so the open port is
|
|
# not an open door.
|
|
honcho_listen_address: "0.0.0.0"
|
|
honcho_listen_port: 8000
|
|
|
|
# Public-facing URL used for absolute links / OIDC callbacks.
|
|
# Leave EMPTY to make Honcho host-agnostic — FastAPI emits relative URLs.
|
|
# Set to a fully-qualified URL only if a specific feature requires it.
|
|
honcho_web_url: ""
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# PostgreSQL configuration
|
|
# ---------------------------------------------------------------------------
|
|
honcho_db_user: honcho
|
|
honcho_db_name: honcho
|
|
# honcho_db_password sourced from vault below
|
|
|
|
# pgvector-enabled Postgres uses the same env vars as the stock image
|
|
honcho_postgres_initdb_args: "--encoding=UTF8 --locale=C"
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# LLM provider configuration
|
|
# ---------------------------------------------------------------------------
|
|
# Supported transports: openai, anthropic, gemini. We start with anthropic
|
|
# and can flip to a local OpenAI-compatible endpoint later by changing
|
|
# these knobs plus the API-key env-var name.
|
|
#
|
|
# Honcho has THREE distinct LLM subsystems, each with its own settings
|
|
# class and env-var prefix:
|
|
#
|
|
# DERIVER — generates observations from messages (background worker)
|
|
# SUMMARY — periodic session summaries
|
|
# DIALECTIC— answers theory-of-mind queries (the `peers/{peer}/chat` API)
|
|
#
|
|
# Honcho's dialectic subsystem uses PER-LEVEL config (minimal/low/medium/
|
|
# high/max) under DIALECTIC_LEVELS__<level>__MODEL_CONFIG__*. Defaults
|
|
# point at OpenAI, so any honcho.dev install without an OpenAI key will
|
|
# fail dialectic queries with "Missing API key for openai model config".
|
|
# We override all five levels to Anthropic below in the API template.
|
|
honcho_llm_transport: "anthropic"
|
|
honcho_deriver_model: "claude-sonnet-4-5"
|
|
honcho_summary_model: "claude-sonnet-4-5"
|
|
honcho_dialectic_model: "claude-sonnet-4-5"
|
|
|
|
# Deriver tuning. Honcho batches representation tasks until the per-batch
|
|
# token threshold is reached. For homelab use with one chatty operator,
|
|
# that means short bursts of conversation can sit unprocessed forever.
|
|
# DERIVER_FLUSH_ENABLED=true bypasses the batching threshold so each
|
|
# message is processed promptly.
|
|
#
|
|
# honcho_deriver_enabled controls whether the deriver Quadlet is created.
|
|
# honcho_deriver_autostart controls whether the service starts on boot.
|
|
# Set autostart=false to create the service but leave it disabled — useful
|
|
# after discovering the deriver burns tokens autonomously despite client-side
|
|
# cadence=0 settings.
|
|
honcho_deriver_enabled: true
|
|
honcho_deriver_autostart: false
|
|
honcho_deriver_workers: 1
|
|
honcho_deriver_polling_seconds: "1.0"
|
|
honcho_deriver_flush_enabled: true
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Embeddings
|
|
# ---------------------------------------------------------------------------
|
|
# Anthropic does not provide an embedding API. Honcho defaults to
|
|
# OpenAI text-embedding-3-small, which fails without an OpenAI key.
|
|
# Three options going forward:
|
|
# 1. Set EMBED_MESSAGES=false — disables semantic search/vector recall
|
|
# but theory-of-mind derivation still works. Default for now.
|
|
# 2. Provide an OpenAI API key purely for embeddings (cheap; embeddings
|
|
# are ~$0.02 per million tokens). Set honcho_embed_messages=true and
|
|
# add vault_honcho_openai_api_key.
|
|
# 3. Stand up a local BGE/nomic embedding endpoint (e.g. on astro-orbiter
|
|
# once GPU is ready) and point Honcho at it via
|
|
# EMBEDDING_MODEL_CONFIG__OVERRIDES__BASE_URL.
|
|
honcho_embed_messages: true
|
|
|
|
# Embedding provider (transport must be openai-compatible). The default
|
|
# OpenAI endpoint requires honcho_openai_api_key. To swap to a local
|
|
# OpenAI-compatible embedder (e.g. Ollama), set honcho_embedding_base_url
|
|
# to the upstream /v1 URL and the API key can be any non-empty string.
|
|
honcho_embedding_transport: "openai"
|
|
honcho_embedding_model: "text-embedding-3-small"
|
|
honcho_embedding_base_url: ""
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Auth
|
|
# ---------------------------------------------------------------------------
|
|
# AUTH_USE_AUTH=true means clients must present a JWT signed with
|
|
# AUTH_JWT_SECRET. The secret lives in vault.
|
|
honcho_auth_enabled: true
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Vault inputs (defined in group_vars/all/vault, encrypted with ansible-vault)
|
|
# ---------------------------------------------------------------------------
|
|
# vault_honcho_database_password - postgres role password
|
|
# vault_honcho_jwt_secret - 32+ byte random string for JWT signing
|
|
# vault_honcho_anthropic_api_key - Anthropic API key for Claude calls
|
|
# vault_honcho_openai_api_key - OpenAI API key, embeddings-only
|
|
honcho_db_password: "{{ vault_honcho_database_password }}"
|
|
honcho_jwt_secret: "{{ vault_honcho_jwt_secret }}"
|
|
honcho_anthropic_api_key: "{{ vault_honcho_anthropic_api_key }}"
|
|
honcho_openai_api_key: "{{ vault_honcho_openai_api_key }}"
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Health check
|
|
# ---------------------------------------------------------------------------
|
|
honcho_health_check_retries: 30
|
|
honcho_health_check_delay: 2
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Quadlet location (rootful)
|
|
# ---------------------------------------------------------------------------
|
|
honcho_quadlet_dir: /etc/containers/systemd
|