f2a7206c7d
fix(security): F01 (Astra P0) — KI-Tool-Ausführung ohne Freigabe verhindern
...
Check Cross-Plugin Imports / check (push) Waiting to run
Vorher: _execute_tool (agent_loop.py) und der KI-Chat-Loop
(stream_chat_comm) führten JEDES im Registry registrierte Tool aus, wenn
das LLM dessen Namen lieferte — ohne Abgleich mit der angebotenen Liste,
ohne required_permission-Check. Reproduktion (Astra): Nur audit_allowed
angeboten, Modell nannte audit_restricted (system:admin) → Handler lief.
Fix (fail-closed, an ALLEN Ausfuehrungspfaden):
- _check_tool_access: (1) Allowlist — nur Tools die dem LLM angeboten
wurden duerfen laufen; (2) required_permission gegen die AKTUELLEN
User-Rechte (deny-first, Rechteentzug wirkt sofort, ohne Kontext =
Ablehnung). Guard vor dry-run/approval/execute-Pfaden.
- stream_chat_comm: gleicher Allowlist-Guard vor execute_tool_call.
- run_react_loop/agent_runner/agent_stream/agent_routes reichen
user_permissions durch (perm_ctx bzw. Session-User).
- check_permission: Session-Kontexte tragen denied_permissions statt
denied — beide Keys werden gelesen, Deny-Liste wird nie mehr ignoriert.
Tests: test_agent_loop.py 18/18 (7 neue F01-Tests nach Astra-Abnahme:
nicht angeboten → Handler null; fehlende Permission → abgewiesen;
Fail-closed ohne Kontext; Deny-Liste session-shape; Rechteentzug
mitten im Lauf wirkt auf naechste Aktion; dry-run guardet auch).
ruff clean. Pre-existing-Beweis: permission_system_live-Failures
reproduzieren sich ohne diesen Patch identisch (Plugin-Aktivierung in
ephemeraler Test-DB, bekanntes Vorbestands-Finding).
2026-09-17 22:48:59 +02:00
335762dd3d
feat(M6): Weitere Hosts — MiniApps in Fenstern + AI-Agenten-Ausgabe-Bloecke ( #364 )
...
- Windows-Host: openMiniAppWindow-Helper + MiniAppWindowContent (windowStore);
Oeffnen-Buttons im Chat-Block (MiniAppBlock) und Dashboard-Widget
- AI-Agenten-Host: Core-Tool send_miniapp (app/ai/miniapp_tools.py) —
miniapp-Block in Agent-Chat (approval_request-Praezedenz), Permission
fail-closed gegen aufrufenden User pro App; Registrierung im lifespan
- agent_loop: tool_context + agent_name (Raum-Aufloesung)
- Fix: MiniAppBlock nutzt useMiniapps (component-Feld) statt Legacy /comm/miniapps
- Tests: M6 7/7 (TDD rot->gruen), Backend-Regression 57/57, Vitest 26/26
(4 neue Window-Tests), tsc clean, build OK
2026-08-31 01:04:33 +02:00
a7699d3598
refactor(block-h): resolve CommConversation hardcoding via kommunikation contract
2026-08-23 12:02:29 +02:00
637cfa7940
refactor(block-h): move AI tool registry into core AI layer
2026-08-23 11:44:14 +02:00
a614ab337b
fix: schema drifts, RLS policies, wiki plugin, agent_loop syntax, test imports, frontend error handling
...
Check Cross-Plugin Imports / check (push) Has been cancelled
- Migration 0135: Fix 3 VARCHAR length drifts + 2 missing tables (forgejo_reported_errors, pgp_keys)
- Migration 0136: Fix 8 RLS policies referencing app.tenant_id instead of app.current_tenant_id
- wiki/__init__.py: Import WikiPlugin for discover_builtins()
- wiki/plugin.py: Fix SyntaxError (unterminated triple-quoted string)
- agent_loop.py: Fix SyntaxError (stray n character in dict)
- test_p1_6_dms_streaming.py: Fix import (CHUNK_SIZE removed, use _sanitize_filename only)
- conftest.py: Use create_all only (alembic conflicts with create_all in tests)
- frontend errorTypes.ts: asError() now handles nested detail objects
- AGENTS.md: Sub-agents forbidden in this project
- DAMAGE_REPORT.md + SCHEMA_DRIFTS.md: Complete damage assessment
- scripts/schema_drift_check.py: Schema drift checker tool
Tests: 24/24 Phase J + 12/12 Phase K = 36/36 passed
tsc: 0 errors
Frontend build: successful
2026-08-21 10:02:50 +02:00
62793a001c
feat: I-WORK-HANDOFF + I-WORK-PROACTIVE — approval requests posted to communication with approval_request block, proactive suggestions posted to communication with action_card block
Check Cross-Plugin Imports / check (push) Has been cancelled
2026-08-21 00:05:13 +02:00
f3fbb5d1e8
fix: remove broken agent_workstream import from agent_loop.py (was deleted in cleanup)
2026-08-19 09:53:12 +02:00
dc1321c78b
feat(I): I-APPR-LOOP — agent loop human-in-the-loop approval integration (require_approval + approval_tools params, ApprovalRequest creation, workstream notification, pause loop), 8 tests passing
2026-08-19 00:20:36 +02:00
638e3f3e1e
feat(F): F-PERM permissions, F-APPR approval, F-AIUSE metadata, F-TRANS transparency, F-DATA-POL data policy, F-OVERSIGHT decision record, F-DRY dry-run, F-AUDIT audit log
...
Check Cross-Plugin Imports / check (push) Has been cancelled
- F-PERM: app/ai/agent_permissions.py (230 lines) — AgentPermissionContext, resolve_agent_permissions(), filter_visible_agents(), check_agent_execute_permission(), optimistic locking
- F-APPR: app/core/approval.py (160 lines) + app/routes/approvals.py (305 lines) + migration 0123 — ApprovalRequest model, CRUD API, approve/reject/expire
- F-AIUSE: app/ai/ai_use_case.py (156 lines) — AIUseCaseMetadata Pydantic model, validate_ai_use_case()
- F-TRANS: app/ai/transparency.py (60 lines) — mark_as_ai_generated(), is_ai_participant()
- F-DATA-POL: app/ai/data_policy.py (210 lines) — enforce_data_policy() with SENSITIVE_FIELDS + provider compliance
- F-OVERSIGHT: app/ai/oversight.py (108 lines) — DecisionRecord, create_decision_record()
- F-DRY: agent_loop.py updated with dry_run parameter
- F-AUDIT: agent_loop.py updated with audit log for tool calls
- agent_routes.py: AI use case metadata endpoints added
- main.py: approval routes registered
- All Python compile checks pass
2026-08-17 16:57:50 +02:00
c760b5961c
feat(F-LOOP): true ReAct loop with structured Thought/Action/Observation step tracking
...
Check Cross-Plugin Imports / check (push) Has been cancelled
- app/ai/agent_loop.py: ReActStep + ReActResult dataclasses, run_react_loop()
with LLM→Tool→Observe loop, max_steps/timeout graceful stop, ErrorCategory
retry (TRANSIENT→retry, PERMANENT→stop, PARTIAL→continue), cost accumulation,
agent.step hook, on_step callback
- app/plugins/builtins/automation/models.py: AgentRunStep model
- alembic/versions/0121_agent_run_steps.py: migration for agent run steps table
- app/plugins/builtins/automation/agent_runner.py: refactored to use run_react_loop(),
saves steps to DB, updates AgentRun with cost/status/duration
- tests/test_agent_loop.py: 11 tests (all passing, mocked, no DB/LLM needed)
- PROGRESS.md: Phase F started, F-LOOP marked done
2026-08-17 16:11:26 +02:00
Agent Zero