AI-powered virtual executive team — a single coherent executive persona backed by 8 specialist agents (FastAPI + Next.js).
Honcho config: default to Anthropic direct, fix ignored table keys
Rufus651d7510a415
diff --git a/docker/honcho/Dockerfile b/docker/honcho/Dockerfileindex b4a539bc..ecb8beec 100644--- a/docker/honcho/Dockerfile+++ b/docker/honcho/Dockerfile@@ -2,7 +2,7 @@## Mirrors plastic-labs/honcho's official Dockerfile (Python 3.13-slim,# uv 0.9.24) but pins the Honcho source to a specific release tag, bakes-# in OE's config.toml (OpenRouter routes every LLM call, dreaming on,+# in OE's config.toml (Anthropic direct for every LLM slot, dreaming on,# AUTH_USE_AUTH=true), and bundles an embeddings sidecar# (`docker/honcho/embed_server.py`) that runs as the `embed` process# serving `BAAI/bge-small-en-v1.5` locally via fastembed/ONNX sodiff --git a/docker/honcho/README.md b/docker/honcho/README.mdindex 7f5e7722..7ecbfd47 100644--- a/docker/honcho/README.md+++ b/docker/honcho/README.md@@ -44,7 +44,7 @@ deriver fail every embed call with a dimension mismatch.|---|---|| `DATABASE_URL` | Postgres connection string || `AUTH_JWT_SECRET` | Every issued `HONCHO_API_KEY` derives from this — store it in a password manager; rotating it invalidates all keys |-| `LLM_OPENAI_API_KEY` | Honcho's `openai` transport accepts any OpenAI-format endpoint, so this can be an OpenRouter key with `base_url` set per slot under `[…model_config.overrides]` in `config.toml` |+| `LLM_ANTHROPIC_API_KEY` | Every LLM slot in `config.toml` uses Honcho's `anthropic` transport. To route through OpenRouter or another OpenAI-compatible endpoint instead, set `LLM_OPENAI_API_KEY`, `LLM_OPENAI_BASE_URL` and the per-slot `<SLOT>_MODEL_CONFIG__TRANSPORT=openai` / `__MODEL=` overrides listed at the top of `config.toml` || `EMBEDDING_MODEL_CONFIG__OVERRIDES__BASE_URL` | Point at the `embed` process, e.g. `http://embed:8001/v1` || `EMBEDDING_MODEL_CONFIG__OVERRIDES__API_KEY` | The sidecar needs no auth, but the OpenAI client requires some string || `EMBEDDING_VECTOR_DIMENSIONS` | `384` for `bge-small-en-v1.5`. Honcho's default schema is `Vector(1536)` |diff --git a/docker/honcho/config.toml b/docker/honcho/config.tomlindex f56d432a..a59cd1a5 100644--- a/docker/honcho/config.toml+++ b/docker/honcho/config.toml@@ -5,16 +5,25 @@# come from environment variables at runtime — see README.md.## Provider strategy:-# - All LLM calls (deriver, summary, dialectic levels, dream deduction)-# use Honcho's `openai` transport but point at OpenRouter. OpenRouter-# fronts Anthropic / OpenAI / Voyage / etc., so a single secret-# (`LLM_OPENAI_API_KEY` = OE's existing OpenRouter key) drives every-# outbound LLM call. The `model` field uses OpenRouter slugs-# (`anthropic/claude-haiku-4.5`, etc.).+# - Every LLM slot (deriver, summary, the dialectic levels, both dream+# passes) uses Honcho's `anthropic` transport with the same Claude+# models Open Executive itself defaults to, so `LLM_ANTHROPIC_API_KEY`+# is the only LLM secret the sidecar needs.+# - To route through OpenRouter (or any OpenAI-compatible endpoint)+# instead, override per slot from the environment; Honcho reads+# `<SLOT>_MODEL_CONFIG__TRANSPORT`, `<SLOT>_MODEL_CONFIG__MODEL` and+# `<SLOT>_MODEL_CONFIG__OVERRIDES__BASE_URL`, e.g.+# DERIVER_MODEL_CONFIG__TRANSPORT=openai+# DERIVER_MODEL_CONFIG__MODEL=anthropic/claude-haiku-4.5+# DIALECTIC_LEVELS__LOW__MODEL_CONFIG__TRANSPORT=openai+# DREAM_DEDUCTION_MODEL_CONFIG__TRANSPORT=openai+# LLM_OPENAI_API_KEY=<key> LLM_OPENAI_BASE_URL=https://openrouter.ai/api/v1+# A `base_url` written directly under `[…model_config]` is silently+# ignored (Honcho ≥ 3.1): in a file it belongs under+# `[…model_config.overrides]`.# - Embeddings come from a local fastembed sidecar (see# docker/honcho/embed_server.py) — no external embedding vendor.-# Quality ladder + the OpenRouter fallback are documented in-# README.md ("Embedding quality ladder").+# Quality ladder is documented in README.md ("Embedding quality").[app]LOG_LEVEL = "INFO"@@ -38,56 +47,39 @@ POOL_SIZE = 10USE_AUTH = true# -------------------------------------------------------------------- #-# LLM routing — everything via OpenRouter using the openai transport.-# The endpoint goes under each slot's `[…model_config.overrides]`: Honcho-# ≥ 3.1 reads `base_url` only from there (a `base_url` directly under-# `model_config` is silently ignored and calls fall through to-# api.openai.com).-# Cheap-tier work (deriver, summary, low-tier dialectic) → Haiku for-# cost; medium+ dialectic and both dream passes → Sonnet for quality.+# LLM routing — Anthropic direct. Cheap-tier work (deriver, summary,+# low-tier dialectic) → Haiku for cost; medium+ dialectic and both dream+# passes → Sonnet for quality. Same model ids as Open Executive's own+# defaults (packages/core/openexecutive/config.py).# -------------------------------------------------------------------- #[deriver.model_config]-transport = "openai"-model = "anthropic/claude-haiku-4.5"-[deriver.model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-haiku-4-5"[summary.model_config]-transport = "openai"-model = "anthropic/claude-haiku-4.5"-[summary.model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-haiku-4-5"[dialectic.levels.minimal.model_config]-transport = "openai"-model = "anthropic/claude-haiku-4.5"-[dialectic.levels.minimal.model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-haiku-4-5"[dialectic.levels.low.model_config]-transport = "openai"-model = "anthropic/claude-haiku-4.5"-[dialectic.levels.low.model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-haiku-4-5"[dialectic.levels.medium.model_config]-transport = "openai"-model = "anthropic/claude-sonnet-4.6"-[dialectic.levels.medium.model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-sonnet-5"[dialectic.levels.high.model_config]-transport = "openai"-model = "anthropic/claude-sonnet-4.6"-[dialectic.levels.high.model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-sonnet-5"[dialectic.levels.max.model_config]-transport = "openai"-model = "anthropic/claude-sonnet-4.6"-[dialectic.levels.max.model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-sonnet-5"[dream]# Dreaming ON. This is Honcho's offline reflection pass — it synthesizes@@ -95,25 +87,20 @@ base_url = "https://openrouter.ai/api/v1"# consistently anxious about cash flow", "the team disagrees on hiring# velocity"). Without dreaming, peer cards are reactive snapshots; the# knowledge-graph value largely lives in this layer. Cost: continuous-# background LLM spend on OpenRouter — see the runbook's cost-monitoring-# section before going to prod.+# background LLM spend — watch it before going to prod.ENABLED = true# Both dream passes. The table names are `deduction_model_config` and# `induction_model_config` (Honcho ≥ 3.1); the older# `[dream.deduction.model_config]` spelling is ignored, which left both-# passes on Honcho's default OpenAI model name.+# passes on Honcho's default model name.[dream.deduction_model_config]-transport = "openai"-model = "anthropic/claude-sonnet-4.6"-[dream.deduction_model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-sonnet-5"[dream.induction_model_config]-transport = "openai"-model = "anthropic/claude-sonnet-4.6"-[dream.induction_model_config.overrides]-base_url = "https://openrouter.ai/api/v1"+transport = "anthropic"+model = "claude-sonnet-5"# -------------------------------------------------------------------- ## Embeddings — local fastembed sidecar (embed_server.py running as the