AI generation
Orvanta provides integrated AI assistance across its platform to enhance the coding and workflow development experience. The system supports multiple AI models and providers, with capabilities spanning code generation, completion, and natural language interaction.
Core features
Section titled “Core features”Orvanta AI chat
Section titled “Orvanta AI chat”The platform includes a comprehensive AI chat interface with two primary modes:
Navigation mode assists with understanding Orvanta itself, including explanations of features, platform navigation, documentation references, API information retrieval, and form auto-population for scripts and flows.
API mode enables direct workspace operations through conversation, covering jobs, scripts, flows, resources, variables, schedules, and workers management.
Script editor AI capabilities
Section titled “Script editor AI capabilities”The integrated chat panel (accessible via Cmd+L on Mac or Ctrl+L on Windows/Linux) supports code generation, issue identification, optimization suggestions, and documentation. When code is selected, it automatically includes context in messages.
Code completion features real-time suggestions during typing, currently requiring “Mistral Codestral through their FIM (Fill-in-Middle) endpoint.” Individual users can toggle completion on or off in settings.
The summary copilot generates descriptive summaries from existing code.
Flow editor AI features
Section titled “Flow editor AI features”The flow editor includes an AI chat panel enabling natural language flow creation and modification. The system can add scripts, loops, branches, search workspace resources and the Orvanta Hub, generate new steps when needed, and automatically configure flow inputs and parameter connections.
Script mode within flow chat provides the same assistance as the script editor. The step input copilot suggests inputs based on previous step results. Iterator expressions for loops and predicate expressions for branches are suggested contextually.
Full-code app builder
Section titled “Full-code app builder”The full-code app builder features an integrated AI chat for natural language app creation and modification, enabling step-by-step application development.
Additional capabilities
Section titled “Additional capabilities”AI form filling automatically populates script and flow run forms based on requirements, with customizable instructions.
CRON schedule generation creates scheduling expressions from natural language prompts.
Supported models
Section titled “Supported models”Orvanta AI integrates with:
- OpenAI models (excluding o1)
- Azure OpenAI
- Anthropic Claude models (including 3.7 Sonnet with extended thinking)
- Mistral Codestral
- DeepSeek
- Google Gemini
- Groq
- OpenRouter
- Together AI
- AWS Bedrock
- Custom OpenAI API-compatible providers (including local Ollama instances)
Custom model names can be specified by typing directly into the model field.
Provider configuration
Section titled “Provider configuration”AWS Bedrock accepts three authentication modes: IAM user keys, bearer API key, or ambient IAM role. The region field is required.
Anthropic offers a “1M context window” toggle enabling extended context on compatible Claude models via the anthropic-beta: context-1m-2025-08-07 header.
Google Gemini and Anthropic on Vertex AI support a platform field switching between standard provider APIs and Google Vertex AI routing.
Custom AI resources accept a headers object for custom header injection, supporting proxies and gateway authentication.
Instance-level configuration
Section titled “Instance-level configuration”Superadmins can configure AI providers, models, and prompts at the instance level for workspace inheritance via Instance settings → AI.
The resolution order prioritizes workspace AI settings when configured, falling back to instance settings otherwise. Workspace members can use AI chat when either scope is configured without requiring settings access.
The AI_HTTP_HEADERS environment variable attaches custom headers to all outgoing AI requests instance-wide, useful for corporate proxies requiring fixed authentication or routing headers.