mirror of
https://github.com/turnstonelabs/turnstone.git
synced 2026-08-12 23:12:23 -06:00
Patrick Buckley
551fc43c15
* feat: per-call model selection on plan_agent / task_agent The calling LLM can now pass `model="<alias>"` to plan_agent or task_agent to override the operator-configured per-kind model for that one invocation. Useful when subtask difficulty varies within a session: the model can downgrade to a cheap alias for trivial work and reach for a stronger one when the problem is hard. Tool descriptions list the live registered aliases (refreshed when the operator hits "sync to nodes" / internal_model_reload), so the calling LLM always sees the current options. Bad aliases return a corrective error dict with the available choices so the LLM retries cleanly rather than failing silently. No whitelist — any alias the registry knows is acceptable; cost control is intentionally ceded to the model. No per-call effort override (out of scope; effort stays operator-configured). Resolution precedence in _run_agent: explicit per-call agent_alias override > registry per-kind (plan_model/task_model) > legacy agent_model > session model. The plan retry path (when _validate_plan fails) reuses the same alias so coaching reflects real model behaviour rather than a different model masking the signal. Implementation: - plan_agent.json / task_agent.json: optional `model` parameter. - ChatSession._validate_agent_model_override extracts and validates the arg; mirrors the existing empty-prompt error pattern. - _prepare_plan / _prepare_task stash the override in item["model_override"]; _exec_* pass it through. - _run_agent gains agent_alias kwarg with defence-in-depth ValueError on unknown alias. - _render_agent_tool_descriptions deep-copies plan/task entries before mutating description so the module-level TOOLS constant stays untouched across sessions; rebuilds the BM25 tool-search index when active so its text matches what the LLM sees. - server._broadcast_agent_tool_schema_refresh walks active workstreams on internal_model_reload so descriptions update without restart. * fix: clarify no-registry placeholder + avoid double BM25 rebuild Addresses Copilot feedback on PR #361. 1. plan_agent.json / task_agent.json placeholder said the parameter falls back to the "operator-configured plan/task model". That text is what no-registry sessions see (registry-bearing sessions get the templated description with the live alias list); for those single-model sessions, omitting the param falls back to the current session model, not an operator-configured one. Reword so the no-registry user gets accurate guidance. 2. _on_mcp_tools_changed already calls _rebuild_tool_search after merging MCP tools. _render_agent_tool_descriptions also rebuilt the BM25 index when active, so the MCP refresh path was rebuilding twice per refresh. Move the BM25 rebuild out of the private render helper into the public refresh_agent_tool_schemas wrapper — _on_mcp_tools_changed keeps calling the render helper directly (no double rebuild), and registry-reload callers go through the wrapper which still keeps the index in sync.
127 lines
4.9 KiB
TOML
127 lines
4.9 KiB
TOML
# turnstone.toml — shared bootstrap configuration
|
|
#
|
|
# This file is read once at startup. Values here are overridden by
|
|
# environment variables, which are in turn overridden by CLI flags.
|
|
#
|
|
# All sections are optional. Missing sections use binary defaults.
|
|
# Config file location precedence:
|
|
# 1. --config flag
|
|
# 2. $TURNSTONE_CONFIG env var
|
|
# 3. ~/.config/turnstone/config.toml
|
|
|
|
# --- LLM API (turnstone, node, eval) ---
|
|
|
|
[api]
|
|
# base_url = "" # API endpoint; empty = binary default
|
|
# api_key = "" # env: OPENAI_API_KEY or ANTHROPIC_API_KEY
|
|
|
|
# --- Default Model (turnstone, node, eval) ---
|
|
|
|
[model]
|
|
# name = "" # Model ID; empty = provider default (gpt-5 / claude-sonnet-4)
|
|
# temperature = 0.0 # 0 = provider default
|
|
# reasoning_effort = "" # "none", "minimal", "low", "medium", "high", "xhigh", "max"
|
|
# context_window = 0 # 0 = auto-detect from provider capabilities
|
|
# max_tokens = 0 # 0 = provider default
|
|
#
|
|
# Sub-agent routing (plan_agent, task_agent tools). Each falls back to
|
|
# agent_model when unset, then to the session model. Use this to point
|
|
# the rare-but-expensive plan agent at a stronger model than the
|
|
# frequent task agent.
|
|
# agent_model = "" # legacy single-knob: both plan and task share this
|
|
# plan_model = "" # plan_agent override (e.g. "claude" for a smart planner)
|
|
# task_model = "" # task_agent override (e.g. "local" for cheap subtasks)
|
|
# plan_effort = "" # reasoning effort for plan_agent (default: "high")
|
|
# task_effort = "" # reasoning effort for task_agent (default: inherit session)
|
|
#
|
|
# At call time, the calling LLM may also pass `model="<alias>"` to
|
|
# plan_agent / task_agent to override these per-invocation. Tool
|
|
# descriptions list available aliases dynamically; bad aliases return
|
|
# an error so the model retries with a valid choice.
|
|
|
|
# --- Named Models (turnstone, node, eval) ---
|
|
# Define model aliases with per-model overrides. Useful for local model
|
|
# servers or mixing providers. Reference by name with --model flag.
|
|
#
|
|
# [models.local]
|
|
# name = "llama-3-70b"
|
|
# provider = "openai"
|
|
# base_url = "http://localhost:8000/v1"
|
|
# context_window = 8192
|
|
#
|
|
# [models.local.capabilities]
|
|
# supports_vision = false
|
|
# supports_web_search = false
|
|
#
|
|
# [models.claude]
|
|
# name = "claude-opus-4-7"
|
|
# provider = "anthropic"
|
|
|
|
# --- Database (turnstone, node, console) ---
|
|
|
|
[database]
|
|
# url = "" # postgres://user:pass@host/db or /path/to.db
|
|
# env: TURNSTONE_DB_URL
|
|
# SSL params (passed through to SQLAlchemy connection):
|
|
# sslmode = "prefer" # disable, allow, prefer, require, verify-ca, verify-full
|
|
# sslrootcert = "" # path to CA cert for verify-ca/verify-full
|
|
# sslcert = "" # path to client cert (mTLS)
|
|
# sslkey = "" # path to client key (mTLS)
|
|
|
|
# --- Auth (node, console) ---
|
|
|
|
[auth]
|
|
# Auth is always enabled. JWT secret is required.
|
|
# jwt_secret = "" # HS256 signing secret (min 32 bytes recommended)
|
|
# env: TURNSTONE_JWT_SECRET
|
|
|
|
# --- Logging (turnstone, node, console) ---
|
|
|
|
[log]
|
|
# level = "" # "debug", "info", "warn", "error"
|
|
# empty = binary default (warn for CLI, info for servers)
|
|
# env: TURNSTONE_LOG_LEVEL
|
|
# json = false # JSON output; auto-enabled when stderr is not a TTY
|
|
|
|
# --- Session (turnstone, node) ---
|
|
|
|
[session]
|
|
# instructions = "" # Default system message
|
|
# compact_max_tokens = 32768 # Max tokens for context compaction summary
|
|
# auto_compact_pct = 0.8 # Trigger compaction at this % of context window
|
|
|
|
# --- Tools (turnstone, node) ---
|
|
|
|
[tools]
|
|
# timeout = 120 # Tool execution timeout in seconds
|
|
# skip_permissions = false # Auto-approve all tool calls
|
|
|
|
# --- Judge (turnstone, node) ---
|
|
|
|
[judge]
|
|
# enabled = true # Enable intent validation
|
|
# confidence_threshold = 0.7 # Minimum confidence for heuristic verdicts
|
|
# output_guard = true # Scan tool output for security signals
|
|
# redact_secrets = true # Redact detected credentials in output
|
|
|
|
# --- Memory (turnstone, node) ---
|
|
|
|
[memory]
|
|
# relevance_k = 5 # Top-K memories for context injection
|
|
# fetch_limit = 50 # Max memories to fetch for ranking
|
|
# max_content = 32768 # Max memory content size in chars
|
|
# nudge_cooldown = 300 # Min seconds between metacognitive nudges
|
|
# nudges = true # Enable memory nudges
|
|
|
|
# --- MCP (turnstone, node) ---
|
|
|
|
[mcp]
|
|
# config_path = "" # Path to MCP servers config file (JSON)
|
|
# refresh_interval = 14400 # Refresh interval in seconds (default: 4h)
|
|
|
|
# --- Server (node, console) ---
|
|
|
|
[server]
|
|
# max_workstreams = 50 # Maximum concurrent workstreams per node
|
|
# env: TURNSTONE_MAX_WORKSTREAMS
|