feat: tool schema optimization, keyword routing, aider_run coding agent

Tool schema optimization (PLAN__Tool_Schema_Optimization.md Phases 1-3):
- model_registry.py: ROLE_DEFAULT_TOOLS — distill gets [], research/coder get
  narrow tool lists by default; applied in get_role_config() when user hasn't
  configured a custom list
- openai_orchestrator.py: keyword routing via narrow_tools_by_keywords() — scans
  user message + last assistant turn; narrows active schemas to matched categories
  only (e.g. "weather" → 3 web tools instead of 69); zero tools sent for pure chat
- openai_orchestrator.py: _get_cached_tools() — module-level schema cache keyed by
  (role, sorted_tool_list, risk_params); eliminates redundant schema rebuilds
- openai_orchestrator.py: _TOOL_SCHEMA_OVERHEAD 3000 → 500 tokens (schemas now
  excluded from the per-call fixed estimate since they're cached separately)
- tools/__init__.py: CATEGORY_TOOL_MAP + _KEYWORD_CATEGORY_MAP + classify_tool_categories()
  + narrow_tools_by_keywords() — the classifier logic lives here so both orchestrators
  can share it

aider_run tool (cortex/tools/aider.py):
- Invokes Aider as a subprocess with --message --yes-always --no-pretty --no-stream
- Project aliases: cortex / aether_api / aether_frontend / aether_container
- Auto-injects OpenRouter API key from Cortex model registry (no ~/.env needed)
- background=True fires async + registers in agent_manager; notify=True sends push
  notification on completion
- admin-only, confirm-required, TOOL_RISK=high
- .gitignore: added .aider.chat.history.md / .aider.input.history / .aider.llm.history

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Scott Idem
2026-06-03 22:39:44 -04:00
parent 29940c299b
commit 29d8aa4aae
6 changed files with 830 additions and 10 deletions

View File

@@ -87,7 +87,13 @@ from tools.git import (
git_log as _git_log,
git_diff as _git_diff,
)
from tools.agents import spawn_agent as _spawn_agent
from tools.agents import (
spawn_agent as _spawn_agent,
agent_status as _agent_status,
agent_list as _agent_list,
agent_cancel as _agent_cancel,
)
from tools.aider import aider_run as _aider_run
from tools.homeassistant import (
ha_get_state as _ha_get_state,
ha_get_states as _ha_get_states,
@@ -114,6 +120,7 @@ import tools.notify as _mod_notify
import tools.agent_notes as _mod_agent_notes
import tools.git as _mod_git
import tools.agents as _mod_agents
import tools.aider as _mod_aider
import tools.homeassistant as _mod_homeassistant
import tools.ae_database as _mod_ae_database
@@ -140,7 +147,7 @@ TOOL_CATEGORIES: dict[str, list[str]] = {
],
"Aether Tasks": ["ae_task_list"],
"Agent Notes": ["agent_notes_read", "agent_notes_write", "agent_notes_append", "agent_notes_clear"],
"Agents": ["spawn_agent"],
"Agents": ["spawn_agent", "agent_status", "agent_list", "agent_cancel", "aider_run"],
"Home Assistant": ["ha_get_state", "ha_get_states", "ha_call_service"],
"Aether Database": ["ae_db_query", "ae_db_describe", "ae_db_show_view"],
}
@@ -207,6 +214,10 @@ _CALLABLES: dict[str, callable] = {
"git_log": _git_log,
"git_diff": _git_diff,
"spawn_agent": _spawn_agent,
"agent_status": _agent_status,
"agent_list": _agent_list,
"agent_cancel": _agent_cancel,
"aider_run": _aider_run,
"ha_get_state": _ha_get_state,
"ha_get_states": _ha_get_states,
"ha_call_service": _ha_call_service,
@@ -230,6 +241,10 @@ TOOL_ROLES: dict[str, str] = {
"file_write": "admin",
"ae_task_list": "admin",
"spawn_agent": "admin",
"agent_status": "user",
"agent_list": "user",
"agent_cancel": "admin",
"aider_run": "admin",
"email_send": "admin",
"nc_talk_send": "admin",
"http_post": "admin",
@@ -251,6 +266,8 @@ CONFIRM_REQUIRED: set[str] = {
"http_post",
"ha_call_service",
"ae_journal_entry_disable", # disables a journal entry — not easily reversed
"agent_cancel", # kills a running background task
"aider_run", # edits files and commits — irreversible without git revert
}
# Security risk ratings — informational for now; will drive auto-allow tiers later.
@@ -348,8 +365,12 @@ TOOL_RISK: dict[str, str] = {
"git_log": "low",
"git_diff": "low",
# Agents — spawning a subprocess with broad permissions is high
# Agents — spawning is high; lifecycle reads are low; cancel is medium (kills a task)
"spawn_agent": "high",
"agent_status": "low",
"agent_list": "low",
"agent_cancel": "medium",
"aider_run": "high",
# Home Assistant — reads are low; controlling physical devices is high
"ha_get_state": "low",
@@ -388,6 +409,7 @@ _ALL_DECLARATIONS: list[types.FunctionDeclaration] = (
+ _mod_ae_tasks.DECLARATIONS
+ _mod_agent_notes.DECLARATIONS
+ _mod_agents.DECLARATIONS
+ _mod_aider.DECLARATIONS
+ _mod_homeassistant.DECLARATIONS
+ _mod_ae_database.DECLARATIONS
)
@@ -554,3 +576,114 @@ def get_openai_tools_for_role(
if tool_list is not None:
allowed &= set(tool_list)
return [t for t in OPENAI_TOOL_SCHEMAS if t["function"]["name"] in allowed]
# ── Keyword-based tool routing ─────────────────────────────────────────────────
# Maps classifier category names → tool names in that category
CATEGORY_TOOL_MAP: dict[str, list[str]] = {
"web": ["web_search", "web_read", "http_fetch"],
"web_post": ["http_post"],
"file": ["project_file_read", "project_file_list", "file_stat", "file_grep",
"file_diff", "file_syntax_check", "file_read", "file_list", "file_write"],
"git": ["git_status", "git_log", "git_diff"],
"system": ["cortex_restart", "cortex_logs", "cortex_status", "cortex_update", "shell_exec"],
"tasks": ["task_list", "task_create", "task_update", "task_complete"],
"cron": ["cron_list", "cron_add", "cron_remove", "cron_toggle"],
"reminders": ["reminders_add", "reminders_list", "reminders_remove", "reminders_clear"],
"scratchpad": ["scratch_read", "scratch_write", "scratch_append", "scratch_clear"],
"ha": ["ha_get_state", "ha_get_states", "ha_call_service"],
"aether": ["ae_journal_list", "ae_journal_search", "ae_journal_entries_list",
"ae_journal_entry_read", "ae_journal_entry_create", "ae_journal_entry_update",
"ae_journal_entry_disable", "ae_journal_entry_append", "ae_journal_entry_prepend"],
"aether_db": ["ae_db_query", "ae_db_describe", "ae_db_show_view"],
"notifications":["web_push", "email_send", "nc_talk_send", "nc_talk_history"],
"agents": ["spawn_agent", "agent_status", "agent_list", "agent_cancel", "aider_run"],
"notes": ["agent_notes_read", "agent_notes_write", "agent_notes_append", "agent_notes_clear"],
"session": ["session_read", "session_search"],
"ae_tasks": ["ae_task_list"],
"claude": ["claude_allow_dir"],
}
_KEYWORD_CATEGORY_MAP: dict[str, list[str]] = {
"web": ["search", "look up", "what is", "who is", "weather", "forecast",
"news", "find on", "google", "website", "article", "research",
"temperature"],
"web_post": ["post to", "send to", "webhook", "trigger webhook"],
"file": ["read file", "show file", "list file", "directory", "grep",
"search in", "find in", "diff", "compare", "syntax check", "open file"],
"git": ["git", "commit", "branch", "pulled", "merged", "repository", "repo"],
"system": ["restart", "update", "status", "logs", "log", "deploy", "run command",
"shell", "is it running", "health"],
"tasks": ["task", "todo", "to-do", "to do", "add task", "create task",
"pending", "what's on my list"],
"cron": ["schedule", "cron", "every day", "every week", "recurring",
"automate", "job"],
"reminders": ["remind", "reminder", "don't forget"],
"scratchpad": ["scratch", "scratchpad", "working note", "jot down", "notepad"],
"ha": ["home assistant", "light", "thermostat", "turn on", "turn off",
"switch", "sensor", "temperature in", "kitchen", "bedroom", "garage"],
"aether": ["journal", "aether journal", "note entry", "log entry",
"search journal", "ae_journal"],
"aether_db": ["database", "query", "sql", "select", "db", "table",
"schema", "maria", "run query"],
"notifications":["notify", "push notification", "send email", "email",
"talk message", "nextcloud"],
"agents": ["spawn", "sub-agent", "delegate", "spawn agent",
"agent status", "agent list", "cancel agent", "background agent",
"aider", "code change", "edit code", "make a change to", "fix the code"],
"notes": ["agent notes", "private notes", "my notes", "agent_notes"],
"session": ["session", "history", "last time", "what did we", "earlier",
"yesterday", "last week", "previously"],
"ae_tasks": ["ae task", "kanban", "board", "ae_task"],
"claude": ["claude allow", "claude directory"],
}
def classify_tool_categories(message: str) -> list[str]:
"""Return category names whose keywords appear in message (case-insensitive).
Empty return means no tool category matched — route as pure chat with zero tool overhead.
"""
low = message.lower()
return [cat for cat, kws in _KEYWORD_CATEGORY_MAP.items() if any(kw in low for kw in kws)]
def narrow_tools_by_keywords(
message: str,
role_tools: list[str] | None,
context_messages: list[dict] | None = None,
) -> list[str]:
"""Narrow the active tool list to categories relevant to this message.
Also scans the last assistant message in context_messages — this catches follow-up
patterns like "yes, please do that" where the tool intent was expressed by the assistant
in the prior turn and the user is simply confirming.
Returns [] if no keywords matched (zero tool overhead).
Returns keyword-matched tools, intersected with role_tools if role_tools is set.
"""
scan_text = message
if context_messages:
for m in reversed(context_messages):
if m.get("role") == "assistant":
scan_text = scan_text + " " + (m.get("content") or "")
break
matched = classify_tool_categories(scan_text)
if not matched:
return []
seen: set[str] = set()
dynamic: list[str] = []
for cat in matched:
for t in CATEGORY_TOOL_MAP.get(cat, []):
if t not in seen:
seen.add(t)
dynamic.append(t)
if role_tools is not None:
role_set = set(role_tools)
dynamic = [t for t in dynamic if t in role_set]
return dynamic