Roadmap (Legacy)
Last updated: July 10, 2026
Version: 0.4.0
Status: All originally planned features shipped ✅ + deep audit remediation ✅ + GitHub OAuth integration ✅ + full Arabic i18n ✅
Tests: ~3,544 root + ~118 package + ~216 TUI
i18n: ~926 keys (EN + AR)
📊 Kanban Overview
Section titled “📊 Kanban Overview”🟢 Done — All Features Shipped
Section titled “🟢 Done — All Features Shipped”| # | Feature | Effort | Impact | Sprint |
|---|---|---|---|---|
| #1 | Web search tool (DuckDuckGo) | 2 days | 🔥🔥🔥 | Sprint 1 |
| #2 | URL content extractor | 1 day | 🔥🔥🔥 | Sprint 1 |
| #3 | Typing indicators | 2 days | 🔥🔥 | Sprint 1 |
| #4 | Slash commands | 1 day | 🔥🔥 | Sprint 1 |
| #5 | Markdown rendering | 1 day | 🔥🔥 | Sprint 1 |
| #6 | Tool output truncation | 2 days | 🔥🔥🔥 | Sprint 2 |
| #7 | File read/write tools | 3 days | 🔥🔥🔥 | Sprint 2 |
| #8 | Automatic retry with backoff | 2 days | 🔥🔥🔥 | Sprint 2 |
| #9 | Conversation summarization | 3 days | 🔥🔥🔥 | Sprint 3 |
| #10 | Inline Python REPL | 4 days | 🔥🔥🔥 | Sprint 3 |
| #11 | Image generation tool | 2 days | 🔥🔥 | Sprint 3 |
| #12 | Agent personality templates | 1 day | 🔥🔥 | Sprint 4 |
| #13 | Rate limit user feedback | 1 day | 🔥🔥 | Sprint 4 |
| #14 | Emoji reactions | 1 day | 🔥 | Sprint 4 |
| #15 | Quick reply buttons | 2 days | 🔥🔥 | Sprint 4 |
| #16 | Graceful error messages | 1 day | 🔥🔥 | Sprint 4 |
| #17 | Session export | 1 day | 🔥 | Sprint 4 |
| #18 | Knowledge graph context | 2-3 weeks | 🔥🔥🔥 | Sprint 5 |
| #19 | Time travel replay | 1-2 weeks | 🔥🔥 | Sprint 5 |
| #20 | IDE integration (MCP server) | 4-6 weeks | 🔥🔥🔥 | Sprint 6 |
| #21 | Voice/multimodal support | 3-4 weeks | 🔥🔥 | Sprint 6 |
🆕 v0.4.0 — GitHub Integration + Arabic i18n + Security Audit (July 2026)
Section titled “🆕 v0.4.0 — GitHub Integration + Arabic i18n + Security Audit (July 2026)”| Feature | Description | Status |
|---|---|---|
| GitHub OAuth Integration | Read-only OAuth flow (no PAT in .env); GitHubClient (REST + GraphQL + pagination); 7 read endpoints (pulls, issues, commits, workflows, branches, releases); repo picker with clone; activity timeline with filter chips; unified workspace | ✅ Done |
| Full Arabic i18n | ~926 translation keys (EN + AR); JS-side t() for Alpine expressions; all templates wired (workspace, agents, swarm, settings, skills, mcp, dashboard); Calibri font; RTL font sizing | ✅ Done |
| Deep Security Audit | ~40 bugs found and fixed: async HITL auto-approve, adapter access control (Discord/Slack), semantic-cache cross-user leak, delegation auth bypass, autoscaler race, SQLite WAL pragmas (14 stores), model registry lock, checkpoint timeout | ✅ Done |
| Unified Workspace | All 3 cards (Project Files, Git Status, GitHub Telemetry) follow one active repo atomically; filesystem autocomplete | ✅ Done |
🆕 Additional Features (Not in Original Plan)
Section titled “🆕 Additional Features (Not in Original Plan)”| Feature | Description | Status |
|---|---|---|
| FTS5 Memory | SQLite full-text search with BM25 ranking | ✅ Done |
| Swarm Manager | Unified in-process + distributed worker orchestration | ✅ Done |
| Web UI Rebuild | 12-tab settings, Alpine.js, provider management | ✅ Done |
| Provider Integration | 9 built-in providers with model discovery | ✅ Done |
| Settings Persistence | SQLite config_store for all settings | ✅ Done |
| Voice Transcription | Telegram voice message STT | ✅ Done |
| Vision Analysis | Image analysis via LLM vision | ✅ Done |
| Role Presets | Swarm worker role presets (orchestrator, backend, frontend, etc.) | ✅ Done |
| Bug Fixes | 13 audit bugs fixed (RBAC, cron, schema, KG, etc.) | ✅ Done |
| Chaos Testing Framework | Failure injection engine with 10 predefined experiments | ✅ Done (Sprint 19) |
| Config Migration UI | Runtime DB schema migration management | ✅ Done (Sprint 19) |
| Load Testing Infrastructure | Locust + k6 test suites with CI integration | ✅ Done (Sprint 19) |
| Adapter Extraction | Clean platform output abstraction (swarm_output.py) | ✅ Done (Sprint 19) |
| WebSocket → SSE HITL | Full HITL support via Server-Sent Events | ✅ Done (Sprint 19) |
🔧 Architecture Remediation (Sprint 8 — June 2026)
Section titled “🔧 Architecture Remediation (Sprint 8 — June 2026)”| Feature | Description | Status |
|---|---|---|
| P0 Bug Fixes | 5 P0 correctness bugs fixed (agent_handler race, Windows code_exec crash, global session messages, config write race, session store deletion) | ✅ Done |
| Dead Code Removal | 6 dead modules deleted (consumer, dispatcher, recovery, checkpoint, stubs, compact_node) | ✅ Done |
| UnifiedToolExecutor | 3 tool registries consolidated onto UnifiedToolExecutor | ✅ Done |
| Service Facade | Zero private attribute access from UI; service layer facade | ✅ Done |
| Unified Session Stores | Merged session stores into a single coherent store layer | ✅ Done |
| HITL Approval UI | Inline approve/deny panel for tiered tool-safety gates | ✅ Done |
| Session History Loading | Browse and load prior conversations | ✅ Done |
| Agents Page | Dedicated agent inspection/control page | ✅ Done |
| UI Bug Fixes | Telemetry dedup, toast null-ref, cost breaker type, swarm logs, init error surfacing | ✅ Done |
| Cross-Platform Hardening | setup.ps1, portable paths, PowerShell completion, env var overrides | ✅ Done |
| RTL / Arabic Completion | Cairo font, i18n system (150+ translations, 71 CSS selectors) | ✅ Done |
🎨 UI Bug Fixes (Sprint 9 — June 2026)
Section titled “🎨 UI Bug Fixes (Sprint 9 — June 2026)”| Feature | Description | Status |
|---|---|---|
| Dark Mode Dropdown Fix | WCAG-compliant dropdown contrast in dark theme | ✅ Done |
| Model Selection Pipeline | Chat-model selector, provider switch on save, SSE passthrough, API key validation | ✅ Done |
| Bilingual Language System | EN/AR toggle, cookie middleware, shared Jinja2Templates, complete i18n | ✅ Done |
📈 Progress Tracking
Section titled “📈 Progress Tracking”By Sprint
Section titled “By Sprint”Sprint 1 (Week 1): ██████████ 5/5 tasks ✅Sprint 2 (Week 2): ██████████ 3/3 tasks ✅Sprint 3 (Week 3): ██████████ 3/3 tasks ✅Sprint 4 (Week 4): ██████████ 6/6 tasks ✅Sprint 5 (Month 2): ██████████ 2/2 tasks ✅Sprint 6 (Month 3): ██████████ 2/2 tasks ✅Sprint 7 (June 2026): ██████████ Web UI rebuild + memory ✅Sprint 8 (June 2026): ██████████ Architecture remediation ✅Sprint 9 (June 2026): ██████████ UI bug fixes + bilingual ✅Sprint 10 (June 2026):██████████ Model picker + tool-call fallback ✅Sprint 11 (June 2026):██████████ Swarm pro-grade overhaul (Phase 1-4) ✅Sprint 12 (June 2026):██████████ Swarm output routing + bugs ✅Sprint 13 (June 2026):██████████ Active Tasks + test fixes ✅Sprint 14 (July 2026):██████████ HITL gates (all platforms) + test isolation ✅Sprint 15 (July 2026):██████████ ConfigStore atomicity + MCP auth/HITL ✅Sprint 16 (July 2026):██████████ Skill checksums + task cancel/retry ✅Sprint 17 (July 2026):██████████ Config reconciliation + engine refactor ✅Overall: 21/21 original features shipped ✅ + 8 remediation sprints complete ✅
Section titled “Overall: 21/21 original features shipped ✅ + 8 remediation sprints complete ✅”🔜 Remediation Audit — Completed Items (Sprints 14–17)
Section titled “🔜 Remediation Audit — Completed Items (Sprints 14–17)”The following items were identified by the post-remediation weak-points audit and completed in Sprints 14-17.
| Priority | Item | Status |
|---|---|---|
| P0 | HITL Approval Gates — Wire HITL into WebUI + all gateway adapter paths (Web, Telegram, Discord, Slack). Fail-closed danger-tool gating with graph interrupt() + swarm bus + MCP classification. | ✅ Done |
| P0 | Test Isolation Fix — Root-caused 36 failing tests. Fixed KAZMA_SECRET env leak (23 failures), handoff cycle detection, workspace singleton pollution. 36→3 failures. | ✅ Done |
| P0 | Hub API Auth — Rewrote hub/api.py to read X-Kazma-Secret with hmac.compare_digest. | ✅ Done |
| P0 | Route Gating — Added auth middleware to /api/sessions, /api/mcp/servers, /api/approve, /api/system/*. | ✅ Done |
| P0 | Docker Bind Fix — Dockerfile CMD changed to --host 0.0.0.0. | ✅ Done |
| P1 | SSRF Validation — Added URL validation to discover/MCP endpoints. | ✅ Done |
| P1 | Active Tasks Tab — In-flight task tracking, non-blocking dispatch, live polling. | ✅ Done |
| P1 | MCP Auth + HITL — Per-server auth (bearer tokens), trust levels, MCP tool classification + HITL gate. | ✅ Done |
| P1 | ConfigStore Atomicity — WAL journaling, busy_timeout, batch transactions, process-wide singleton, YAML→SQLite reconciliation. | ✅ Done |
| P1 | Skill Checksums — Fail-closed verification, HMAC-SHA256 signatures, kazma hub sign CLI. | ✅ Done |
| P2 | Circuit Breaker Badges — Live per-worker ⚡ breaker state in swarm panel. | ✅ Done |
| P2 | Per-Worker Start/Stop — Individual worker lifecycle control (API + UI). | ✅ Done |
| P2 | Task Cancel/Retry — Cancel running tasks, retry failed tasks from UI. | ✅ Done |
| P2 | Docs Accuracy — Test count, Slack description, README feature descriptions updated. | ✅ Done |
| P2 | Config Reconciliation — YAML auto-seeds SQLite on startup, non-clobbering. | ✅ Done |
| P2 | Engine Refactor — 1,878→1,573 lines. 3 modules extracted (ReliabilityRegistry, WorkerPhonebook, CheckpointManager). | ✅ Done |
Remaining Open Items
Section titled “Remaining Open Items”| Priority | Item | Effort | Status |
|---|---|---|---|
| P2 | Unify routing algorithms (merge 3 implementations) | M | ✅ Done |
| P2 | Semantic routing (embeddings-based capability matching) | L | ✅ Done |
| P2 | Visual pipeline editor (drag-and-drop DAG) | XL |
🏷️ Label Legend
Section titled “🏷️ Label Legend”| Label | Meaning |
|---|---|
sprint-1 through sprint-9 | Which sprint the feature belongs to |
quick-win | High impact, low effort (1-2 days) |
high-impact | Major user value |
polish | UX improvement |
competitive | Catch up to market leaders |
tool | New agent tool |
gateway | Gateway/adapter feature |
memory | Memory/context feature |
core | Core framework feature |
P0 / P1 / P2 | Remediation priority (critical / high / medium) |