Roadmap
What's built, what's in progress, and what's planned. Links point to GitHub issues and pull requests for the latest status.
Built
| Feature | Issues / PRs | Details |
|---|---|---|
| LLM Gateway | — | 25 RPCs — threads, messages, MCP tools, memories, voice |
| Socayo Gateway | — | 6 RPCs — user profiles, coaching plan generation |
| Storage Gateway | — | 13 RPCs — file upload/download, folders, quota |
| Integrations Gateway | — | 12 RPCs — Pipedream Connect, OAuth flows, actions, triggers |
| Webhook Gateway | #226 | 9 RPCs — Convoy integration, endpoints, subscriptions, deliveries |
| Notification Gateway | #636 | 30 RPCs — Novu integration, push, in-app, email, SMS, topics, digest |
| Notification Admin | #636 | 19 RPCs — tenant orgs, providers, workflows, promotion |
| Auth Gateway | #648 | 7 RPCs — API key CRUD, rotation, usage stats, rate limits |
| Streaming LLM Responses | #159 | Streaming via Kafka chunk publishing, reduced perceived latency |
| Context Management | #153 | Windowing and compaction strategies for conversation history |
| MCP Tool Annotations | #157 | Proto annotation-driven MCP tool exposure and registry generation |
| Mem0 Memory Integration | #167 | Self-hosted Mem0 server, semantic memory search and CRUD |
| Workflow Turn Context | — | Turn context injection into LLM generation system prompts |
| Image Generation | — | Image generation via OpenRouter with content part handling |
| Message Timestamps | #720 | Timestamp population on all conversation messages |
| File Metadata Awareness | #166 | File metadata CRUD, tags, and description support |
| Model Fallbacks & Structured Outputs | #851 | Fallback model arrays, JSON schema validation, response healing |
| Voice AI (Pipecat) | — | Real-time voice sessions via Daily + Cerebrium |
| Flux GitOps | — | Automated image deployment, Kyverno policies |
| LGTM Observability | — | Loki, Grafana, Tempo, Mimir + OTel Collector |
| Multi-Tenant Isolation (Phase 4) | #1084 | Tenant-scoped Pinecone namespaces, Mem0 app_id validation, cache key isolation |
| Tenant-Prefixed Actor Keys | #1019 | All Restate actor keys prefixed with tenant ID for state isolation |
| Gateway Identity Enforcement | #1082 | KrakenD rejects org-less JWTs (403), JWT firebase.tenant claim required |
| MCP Tenant Enforcement | #1088 | All MCP tool calls require tenant context |
| API Key Rotation | #992 | RotateApiKey RPC with configurable grace period |
| Rate Limit Headers | #987 | X-RateLimit-* response headers on all API responses |
| Proto Decoder Tool | #1037 | Web UI for decoding protobuf binary to JSON |
| Grafana Dashboards | #1014 | Restate services and application logs dashboards |
| Mem0 Embedding Cache | #1103 | LRU cache for embedding vectors, reducing SearchMemories latency |
| OIDC Multi-Provider Support | #1079 | Publishable keys support Auth0, Okta, AWS Cognito, Keycloak alongside Firebase |
| RPC Error Payload Standards | #1205 | Structured RpcError with ErrorCode enum, ErrorDetails, terminal/retryable classification |
| Memory Timestamp Injection | #1240 | Injected memories include (remembered: <date>) annotations for temporal awareness |
| ConfigureProvider Upsert | #1212 | Provider configuration now updates existing integrations instead of creating duplicates |
| Generation Cancellation Fix | #1231 | Graceful workflow cancellation during tool execution without panics |
In Progress
Features with active development — open issues or pull requests.
| Feature | Issues / PRs | Details |
|---|---|---|
| Documentation Site | #892 | This site (docs.yocaso.dev) |
| Notification Service (remaining) | #691–#693 | Gateway deployed, 42 KrakenD endpoints live, BDD tests in progress; observability pending (open tickets) |
| Fine-Grained Authorization | #649, #650 | OpenFGA service built; gateway-level enforcement not yet wired |
| LLM Web Search | #252, #631 | Basic web search via OpenRouter; deep research (Perplexity Sonar) not yet integrated |
| Convoy Webhook (remaining) | #300–#303, #616 | Core RPCs done; rate limiting, K8s, observability, and Kafka source pending (open tickets) |
| Mem0 Integration (remaining) | #506–#528, #821 | Core CRUD done; batch ops, entities, events, MCP tool exposure, and perf tuning pending (open tickets) |
| Streaming Responses (clients) | #411–#416 | Backend streaming done; mobile/web and Pipecat client handlers pending |
| KrakenD API Gateway | #163 | KrakenD deployed; proto-driven config auto-generation not yet built |
Planned
Features with a PRD or issue but no implementation started.
P1 — High Priority
| Feature | Issues | Details |
|---|---|---|
| Conversation Revert & Branch | #162 | Revert to previous message, branch conversations |
| Cancel Generation | #164 | User-facing explicit generation cancellation (backend cancellation fixed in #1231, UI pending) |
| LLM Feedback & Rating | #161 | Message-level feedback, thumbs up/down, ratings |
| Synchronous Conversation Flow | #857 | Synchronous request/response for LLM conversations |
| Reasoning Tokens (Thinking) | #861 | Extended thinking / reasoning token support |
| Publishable Keys (OIDC) | #774 | Client-side publishable keys with OIDC validation (multi-provider support shipped in #1079, iOS integration pending) |
P2 — Medium Priority
| Feature | Issues | Details |
|---|---|---|
| Time-Aware LLM | #165 | Current time injection, timestamp-aware context |
| FILE_ID Resolution | #158 | URL caching and resolution for file content parts |
| Scheduled Jobs | #253 | Scheduled LLM conversations via Cloud Scheduler |
| Agentic Task Management | #227, #571 | LLM-managed tasks within conversations |
| Agent-Driven Workflows | #228, #546 | LLM-orchestrated workflow execution / BPMN |
| USDA FoodData Central | #635 | Nutritional data lookup for health coaching |
| Apple HealthKit Integration | — | Health data ingestion from iOS devices |
| GCP Events Pipeline | #565 | Firebase Auth, Storage, Firestore, Scheduler events |
| Migrate Pipecat to GCP | #853 | Move voice agent from Cerebrium to in-house GCP |
| Pipecat Voice Tuning | #858 | Fix premature cutoff and tune voice parameters |
| Self-Hosted STT | #859 | Self-hosted speech-to-text (post-GCP migration) |
| Self-Hosted TTS | #860 | Self-hosted text-to-speech (post-GCP migration) |
| Full Plan Domain | #441 | Modular health coaching plans |
P3 — Lower Priority
| Feature | Issues | Details |
|---|---|---|
| Multi-Agent Orchestration | #383, #572 | Conversation router, multi-agent coordination |
| Omnichannel Chat (Chatwoot) | #637, #652 | Multi-channel messaging via Chatwoot |
| Browser Automation | #254 | Remote browser automation for AI agents |
| Agentic Payments (iOS) | #384 | Conversational checkout with Stripe |
| PostHog Analytics | #443 | Product analytics and experimentation |
| BigQuery Data Lake | #428 | Kafka Connect + dbt data pipeline |
Enterprise Readiness
| Feature | Issues | Details |
|---|---|---|
| Enterprise Readiness | #451 | Umbrella: SSO, audit, billing, compliance |
| Enterprise Developer Portal | #541 | Developer portal for API consumers |
| End-to-End Multi-Tenancy | #539 | Core isolation done (Phase 4); RTDB restructuring and OpenFGA enforcement remaining |
| SSO & Directory Sync | #536 | Ory Polis / BoxyHQ integration |
| Audit Logging | #537 | Enterprise audit trail via Retraced |
| SOC 2 Compliance | #538 | Compliance automation with Comp AI |
| Billing & Metering | #535 | Usage metering and invoicing via Lago |
| VPN-Only Internal Endpoints | #854 | Restrict internal endpoints to VPN access |
POC / Exploration
| Feature | Issues | Details |
|---|---|---|
| React-Admin Portal | #586 | Enterprise admin portal POC |
| Status Page | #579 | Uptime Kuma status page |
| Dashboard Embedding | #721 | Superset + Cube embedding POC |