mirror of
https://github.com/openswarm-ai/openswarm.git
synced 2026-09-26 19:44:51 +02:00
[arnav] remove dead code identified by audit
Each removal verified by checking actual production callers (frontend,
electron, internal HTTP, MCP-server subprocesses) — not just test
references. Symbols whose only callers were tests are removed along
with those tests.
Production removals (~390 LOC):
- backend/main.py
- websocket_session: drop `agent:edit_message` WS branch. Frontend
only ever uses HTTP `POST /api/agents/sessions/{id}/edit_message`
(frontend/src/shared/state/agentsSlice.ts); nothing on the wire
sends a WS message of this type.
- backend/apps/agents/agent_manager.py
- AgentManager._build_connected_tools_context (~80 LOC): zero call
sites in production; the connected-tools system-prompt context is
built inline in _compose_system_prompt now.
- AgentManager._approx_tokens / _summarize_message_block: pure
helpers whose only callers were tests. The compaction path uses
LLM-driven _maybe_compact instead.
- backend/apps/agents/browser_agent.py
- clear_browser_history: only used by tests. _browser_history is
pruned via the size cap inline.
- MODEL_MAP constant: never read.
- backend/apps/agents/mcp_preflight.py
- DISCOVERY_SCAFFOLDING (~25-line system-prompt block): defined but
never appended anywhere. The header comment described an intended
use that the codebase no longer has.
- backend/apps/agents/providers/registry.py
- thinking_params_for, _is_9router_available, OPENROUTER_BASE_URL,
get_context_window: zero callers in production. Thinking-params
routing is done by the provider classes directly; 9Router presence
is detected at request time; context-window numbers are stamped
onto sessions from BUILTIN_MODELS at launch.
- backend/apps/agents/tools/{base,web}.py
- BaseTool.get_schema (abstract) + WebSearchTool/WebFetchTool
overrides: production code in backend/apps/web/web.py instantiates
these tools and only calls .execute(); the JSON-schema lives in
the HTTP wrapper, not on the tool class.
- backend/apps/outputs/outputs.py
- _resolve_model + MODEL_MAP: tests-only.
- load_output: docstring claimed it was a public helper for "other
modules" but no module imported it.
- backend/apps/service/client.py
- set_user_id, the _user_id module global, and the dead cache short-
circuit in _get_user_id: setter was tests-only. _get_user_id now
reads user_email directly from settings on every call.
- backend/apps/settings/credentials.py
- get_provider_credentials: zero callers. The sibling get_agent_sdk_env
is kept (it has the explicit "Legacy helpers" keep-comment).
Test updates:
- test_agent_manager_unit.py: drop _approx_tokens / _summarize_message_block
cases (5 tests), update module docstring index.
- test_browser_agent_unit.py: drop clear_browser_history cases (2 tests)
and the unused _Boom helper class in the repr-fallback test.
- test_outputs_unit.py: drop _resolve_model / load_output cases
(4 tests), update docstring + import list.
- test_v2_invariants.py: drop get_context_window tests + get_schema
assertions on web tools (kept name + BaseTool inheritance checks).
- test_service.py: rewrite the 4 set_user_id-driven tests to drive
user_id through settings.user_email instead, so _get_user_id's live
envelope-stamping path stays covered.
Verification:
- ruff --select F401,F811,F841 backend/ → clean.
- pytest backend/tests/ → 1167 passed, 1 deselected (pre-existing
sandbox git test, unrelated). No tests dropped silently — every
deletion is paired with the corresponding test removal/rewrite.
- Dead-code scan re-run: dead WS events 1→0, Tier-2 high-confidence
14→11 (residue is SDK-callback `context` params + Pydantic `cls`
validators — both false positives vulture can't see through),
vulture total 165→145.
Total diff: -565 / +34 LOC across 15 files.
Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
Arnav Naval
@@ -450,86 +450,6 @@ class AgentManager:
|
||||
logger.info(f"[MCP-DEBUG] Final mcp_servers: {list(mcp_servers.keys())}")
|
||||
return mcp_servers
|
||||
|
||||
def _build_connected_tools_context(self, allowed_tools: list[str]) -> str | None:
|
||||
"""Build a context block describing connected MCP tools and their accounts.
|
||||
|
||||
Tools set to 'deny' and fully-denied servers are excluded.
|
||||
"""
|
||||
all_tools = load_all_tools()
|
||||
mcp_tools = [t for t in all_tools if t.mcp_config and t.enabled and t.auth_status in ("configured", "connected")]
|
||||
|
||||
sections = []
|
||||
for tool in mcp_tools:
|
||||
tool_ref = f"mcp:{tool.name}"
|
||||
if tool_ref not in allowed_tools and allowed_tools != get_all_tool_names():
|
||||
continue
|
||||
|
||||
if _is_fully_denied(tool):
|
||||
continue
|
||||
|
||||
server_name = _sanitize_server_name(tool.name)
|
||||
denied = _get_denied_tool_names(tool)
|
||||
tool_descs = {
|
||||
k: v for k, v in tool.tool_permissions.get("_tool_descriptions", {}).items()
|
||||
if k not in denied
|
||||
}
|
||||
if not tool_descs:
|
||||
continue
|
||||
|
||||
lines = [f"MCP Server: {server_name}"]
|
||||
lines.append(f" Status: {tool.auth_status}")
|
||||
|
||||
if tool.connected_account_email:
|
||||
lines.append(f" Connected account: {tool.connected_account_email}")
|
||||
lines.append(
|
||||
f" IMPORTANT: When calling tools from this server that require an email "
|
||||
f"parameter (e.g. user_google_email, user_email), always use "
|
||||
f"\"{tool.connected_account_email}\" automatically — do NOT ask the user."
|
||||
)
|
||||
|
||||
# Discord guild scoping — hard restriction. The bot may technically
|
||||
# be in other servers (across other OpenSwarm users), but this
|
||||
# specific user only authorized these guild IDs.
|
||||
if tool.name.lower() == "discord":
|
||||
guilds = tool.oauth_tokens.get("guilds") or []
|
||||
if guilds:
|
||||
guild_descriptions = ", ".join(
|
||||
f"{g.get('name', 'Unknown')} ({g.get('id', '')})" for g in guilds
|
||||
)
|
||||
allowed_ids = [g.get("id", "") for g in guilds if g.get("id")]
|
||||
lines.append(
|
||||
f" AUTHORIZED DISCORD SERVERS (guild_ids): {guild_descriptions}"
|
||||
)
|
||||
lines.append(
|
||||
f" HARD RESTRICTION: You MUST only call Discord tools that operate on "
|
||||
f"these guild_ids: {allowed_ids}. NEVER call Discord tools on any other "
|
||||
f"guild_id even if the bot has access to it. NEVER list, search, or "
|
||||
f"enumerate servers outside this list. If a user asks about a server "
|
||||
f"not in this list, refuse and tell them to authorize it via the Connect "
|
||||
f"Discord button. This is a security boundary, not a preference."
|
||||
)
|
||||
else:
|
||||
lines.append(
|
||||
f" No Discord servers authorized yet. Tell the user to click "
|
||||
f"'Connect Discord' to add a server before attempting any Discord actions."
|
||||
)
|
||||
|
||||
tool_names = list(tool_descs.keys())
|
||||
if tool_names:
|
||||
lines.append(f" Available tools ({len(tool_names)}): {', '.join(tool_names)}")
|
||||
|
||||
sections.append("\n".join(lines))
|
||||
|
||||
if not sections:
|
||||
return None
|
||||
return (
|
||||
"<connected_mcp_tools>\n"
|
||||
"The following MCP tool servers are connected and available. "
|
||||
"Use them directly when relevant to the user's request.\n\n"
|
||||
+ "\n\n".join(sections)
|
||||
+ "\n</connected_mcp_tools>"
|
||||
)
|
||||
|
||||
def _build_outputs_context(self, active_outputs: list[str] | None = None) -> str | None:
|
||||
"""Outputs context for the system prompt.
|
||||
|
||||
@@ -962,86 +882,6 @@ class AgentManager:
|
||||
return ""
|
||||
return "<prior_conversation>\n" + "\n".join(lines) + "\n</prior_conversation>"
|
||||
|
||||
# ------------------------------------------------------------------
|
||||
# Compaction & token guard (Phase 2)
|
||||
#
|
||||
# Triggered by *live* context-usage ratio, not turn count. The signal
|
||||
# is the same `ctx_used_pct` we already broadcast to the UI on every
|
||||
# turn: input_tokens / context_window. Three escalating thresholds:
|
||||
# - compact_threshold_pct (default 0.65): summarize stale tool_results
|
||||
# and old user/assistant pairs before the next query() call
|
||||
# - context_soft_cap_pct (default 0.90): pre-send hard guard. After
|
||||
# compaction, if still over, LRU-trim active_outputs/active_mcps
|
||||
# - >= 1.0 hits the proxy/Anthropic 200K ceiling — friendly card
|
||||
# surfaces from the catch-all
|
||||
# ------------------------------------------------------------------
|
||||
|
||||
@staticmethod
|
||||
def _approx_tokens(text: str) -> int:
|
||||
"""Conservative chars/4 estimate. Used for the pre-send guard
|
||||
and the compaction trigger when a precise count_tokens isn't
|
||||
cheap (or the route isn't Anthropic). Errs slightly high so we
|
||||
compact a touch earlier than strictly necessary."""
|
||||
return max(1, len(text or "") // 4)
|
||||
|
||||
@staticmethod
|
||||
def _summarize_message_block(messages: list) -> str:
|
||||
"""Programmatic, no-LLM summary of a message slice. Mirrors the
|
||||
shape of browser_agent._summarize_messages: extracts the original
|
||||
user task, counts tool calls, captures the last assistant text.
|
||||
Cheap, deterministic, and never makes a network call — so
|
||||
compaction itself adds zero latency to the user's turn.
|
||||
"""
|
||||
if not messages:
|
||||
return ""
|
||||
|
||||
initial_task = ""
|
||||
for m in messages:
|
||||
if getattr(m, "role", "") == "user":
|
||||
content = getattr(m, "content", "")
|
||||
txt = content if isinstance(content, str) else str(content)
|
||||
if txt.strip():
|
||||
initial_task = txt.strip()[:400]
|
||||
break
|
||||
|
||||
tool_calls_by_name: dict[str, int] = {}
|
||||
last_tool_results = 0
|
||||
last_assistant_text = ""
|
||||
for m in messages:
|
||||
role = getattr(m, "role", "")
|
||||
if role == "tool_call":
|
||||
content = getattr(m, "content", {}) or {}
|
||||
name = (content.get("tool") if isinstance(content, dict) else None) or "unknown"
|
||||
tool_calls_by_name[name] = tool_calls_by_name.get(name, 0) + 1
|
||||
elif role == "tool_result":
|
||||
last_tool_results += 1
|
||||
elif role == "assistant":
|
||||
content = getattr(m, "content", "")
|
||||
if isinstance(content, str) and content.strip():
|
||||
last_assistant_text = content.strip()
|
||||
elif isinstance(content, list):
|
||||
for block in content:
|
||||
if isinstance(block, dict) and block.get("type") == "text":
|
||||
txt = (block.get("text") or "").strip()
|
||||
if txt:
|
||||
last_assistant_text = txt
|
||||
|
||||
parts = ["<compacted_history>"]
|
||||
parts.append("[The following is a programmatic summary of earlier turns in this session. Originals are preserved on disk and viewable via the chat UI's compaction drawer.]")
|
||||
if initial_task:
|
||||
parts.append(f'Initial user request: "{initial_task}"')
|
||||
if tool_calls_by_name:
|
||||
total = sum(tool_calls_by_name.values())
|
||||
top = sorted(tool_calls_by_name.items(), key=lambda kv: -kv[1])[:8]
|
||||
parts.append(f"Tool calls so far ({total} total): " + ", ".join(f"{n}×{c}" for n, c in top))
|
||||
if last_tool_results:
|
||||
parts.append(f"Tool results received: {last_tool_results}")
|
||||
if last_assistant_text:
|
||||
parts.append("Last assistant message:")
|
||||
parts.append(last_assistant_text[:1200])
|
||||
parts.append("</compacted_history>")
|
||||
return "\n".join(parts)
|
||||
|
||||
def _maybe_compact(self, session: AgentSession, force: bool = False) -> bool:
|
||||
"""Run summarizer when ctx_used_pct >= compact_threshold_pct (or force).
|
||||
|
||||
|
||||
Reference in New Issue
Block a user