Capabilities
What NovaServe ships today
44 of 44 capabilities are live in the product today; anything still rolling out is marked as such.
Assistant
| Capability | Status | How we do it |
|---|---|---|
| Multi-model chat in one place | Whole NovaServe roster plus routing tiers | |
| Saved conversation threads | Per-thread model and retrieval settings | |
| Document chat / attachments | Attach text documents to any turn | |
| Retrieval-grounded answers | Pick a knowledge base per thread | |
| Scheduled prompts / tasks | Hourly to monthly with full run history | |
| Image, video and audio generation | Image, video and speech models in the studio | |
| Voice mode | Talk to the assistant; spoken replies, metered per minute |
Agents
| Capability | Status | How we do it |
|---|---|---|
| Multi-step tool-calling agent | Nexus agent with real tool execution | |
| Reusable skill library | System skill library, pin any of them | |
| Custom skill builder | Cloud runtime or sandboxed browser worker | |
| Run history and traces | Every tool call logged per run | |
| Deep research agent | Plan, parallel investigation and synthesised brief | |
| Parallel agent fan-out | Research angles run concurrently on the fleet | |
| Browser automation | Stateful browsing tool: open, click, submit, read |
Knowledge / RAG
| Capability | Status | How we do it |
|---|---|---|
| Document retrievers | Chunking, embedding and semantic search | |
| Configurable chunking | Chunk size and overlap per base | |
| Retrieval playground | See scored matches before you ship | |
| Vector store included | No third-party vector DB bill |
Connectors and interoperability
| Capability | Status | How we do it |
|---|---|---|
| Native SaaS connectors | Storage, CRM, data and code providers | |
| MCP server registry | Bring any Model Context Protocol server | |
| Outbound webhooks | Deliver task results anywhere | |
| OpenAI-compatible REST API | Drop-in /v1 endpoints |
Billing and metering
| Capability | Status | How we do it |
|---|---|---|
| Usage-based credits | Per-request metering with clear 402s | |
| Exactly-once billing on retries | Idempotency keys plus resumable streams | |
| Auditable credit ledger | Charged vs consumed on every request | |
| Nightly reconciliation | Ledger vs telemetry mismatch alerts | |
| Shareable, permissioned reports | CSV, PDF and ZIP audit bundles |
Observability and governance
| Capability | Status | How we do it |
|---|---|---|
| Request logs with latency and traces | Per-endpoint admin dashboard | |
| Role-based access control | Server-side role checks everywhere | |
| API key management | Hashed keys, scopes and rate limits | |
| Audit trail of report access | Every shared link open is recorded | |
| SSO / SAML | Domain-based SAML SSO plus Google and email |
Enterprise ML
| Capability | Status | How we do it |
|---|---|---|
| Hosted inference at scale | Streaming inference behind one metered API | |
| Cost-per-credit analytics | Cost per credit by model and endpoint | |
| Feature store with online serving | Register feature groups and flip on low-latency lookups | |
| AutoML with model leaderboard | Task-aware search across six candidate families | |
| Drift monitoring and alerts | PSI, KL, JS or accuracy delta with thresholds |
App building and developer surfaces
| Capability | Status | How we do it |
|---|---|---|
| Prompt-to-app builder | Generated pages, data model and integrations | |
| One-click hosting for generated apps | Published to a NovaServe URL with deploy history | |
| CLI | Chat, images, skills and usage from the terminal | |
| Python and TypeScript SDKs | Streaming clients over the same metered API | |
| Desktop app | Installable progressive web app |
Pricing
| Capability | Status | How we do it |
|---|---|---|
| Published per-model pricing | Published rates for every model we serve | |
| No seat minimums | Pay for the tokens you use |