Roadmap
Hrida AI Roadmap
What we're shipping and what's coming next. Priorities are updated quarterly based on enterprise customer feedback and the evolving AI landscape.
Now — Active Development
Native app for Windows, macOS, and Linux. No Docker, no terminal required. Global hotkey access, system tray integration, instant startup. Installs in one click and manages the backend automatically.
Multi-step agent execution with tool calling, loop detection, and execution tracing. Agents can read and write files, query APIs, and maintain context across long task chains.
Hybrid BM25 + semantic retrieval, reranking, configurable chunking strategies, metadata filtering, and better citation formatting in responses.
SAML 2.0, OpenID Connect, and SCIM 2.0 user provisioning. Group-to-role sync from your identity provider. Just-in-time provisioning on first login.
Per-user and per-model usage metrics, token consumption reporting, cost allocation by team or project, and exportable CSV/API access for billing integration.
Next — Planned
Curated, enterprise-vetted plugin catalog with private registry support for distributing internal tools across your organization.
Full-duplex voice interface with support for local TTS/STT models. Configurable wake word, push-to-talk, and always-listening modes. Works fully offline.
Built-in Model Context Protocol server management. Install, configure, and route tool calls to MCP servers directly from the UI without additional proxies.
Structured document collections with version history, shared annotations, and AI-assisted editing. Distinct from the existing Knowledge base — a workspace-level document layer.
Route requests automatically based on cost, latency, capabilities, or custom policy. Define fallback chains for high-availability deployments across multiple providers.
Later — In Scope
Native support for active-active multi-region setups with regional data residency controls. Suitable for global enterprises with GDPR and data localization requirements.
LoRA and QLoRA fine-tuning directly from the admin UI using conversation history as training data. Export and deploy fine-tuned models back into the platform's model registry.
Configurable policy layers per model and per user group. Prompt injection detection, PII redaction in transit, output filtering, and audit logging for compliance workflows.
iOS and Android apps that connect to your self-hosted instance. End-to-end encrypted sync, biometric auth, offline caching, and push notifications for async completions.
Supervisor–worker agent topologies. Parallel agent execution with shared state, cross-agent tool sharing, and visual workflow editor for non-technical orchestration.
Internal registry for versioning, tagging, and deploying custom and fine-tuned models with approval workflows and rollback support.
Built-in benchmark suite for evaluating model responses against enterprise-defined quality criteria. Red-teaming tooling for adversarial prompt testing before production rollout.
Native connectors for SharePoint, Confluence, Salesforce, ServiceNow, and enterprise databases — sourcing live context for RAG without manual document uploads.
Shipped — Recent Releases
See the Changelog for the complete release history.
Notable recent milestones:
- Multi-model routing — simultaneous access to Ollama, OpenAI, Azure OpenAI, Anthropic, Bedrock, and Gemini from a single interface
- Hrida Terminal — fully integrated browser-based terminal for server-side shell access and model CLI tools
- RAG knowledge base — collection-level document stores with hybrid retrieval and per-conversation knowledge scoping
- Tool calling & Functions — Python function definitions exposed as callable tools within conversations
- RBAC & workspace isolation — group-level permissions, workspace-scoped models, and user provisioning API
- Pipeline framework — programmable middleware for request/response transformation, filtering, and enrichment
This roadmap reflects current priorities and is subject to change. Enterprise customers can influence prioritization directly — contact sales or reach us at hello@hrida.ai. Feature requests and community discussion: GitHub Discussions.