- Public release
- Add support for self-hosting
- Workspace wide API keys, cost tracking and usage limits
- Workspace roles (owner vs member)
- Markdown rendering for run responses
- Paging, filtering and CSV export for endpoint logs
- Add support for Mistral AI (Small, Large)
- Add support for Anthropic Claude 3 (Sonnet, Opus)
- Add support for system prompt in Anthropic models
- Add support for Mistral embedding model (mistral-embed)
- Support for latest GPT-3.5 Turbo and GPT-4 Turbo Preview models (0125)
- Support for new OpenAI embedding models (text-embedding-3-small, text-embedding-3-large)
- Increased prompts per project limit from 100 to 250
- Linear integration
- Support for Meta Llama 2 through Hugging Face
- Reusable test data
- Cancel long running LLM requests
- Sort and filter responses
- Improved layout for prompts
- Support for Gemini Pro (preview)
- Improved text editor performance and usability
- Show token counts in responses
- Collapsible sidebar
- Improved save prompt behaviour
- GitHub integration
- Rating prediction
- Generate prompt improvement suggestions
- Specify persona to automate testing
- Ability to loop a branch in a chain
- Trigger manual interrupt from code block
- Specify a reason for response ratings
- Support for seed and JSON mode in OpenAI models
- Improved visualisation of chain runs
- Project specific API keys
- Daily cost tracking per model
- Specify usage limits with notifications
- Project roles (owner vs member)
- Improved logs for multi-step requests
- Documentation
- Rate run responses
- Support for PaLM 2 for Chat
- Improved version history
- Improved test data entry
- Test function interrupts dynamically in the UI
- Support for GPT-4 Turbo (preview) and new GPT-3.5 Turbo
- Save default model and parameters
- Onboarding questionnaire
- Support for branching in chains
- Test chat conversations in the UI
- Improved admin dashboards
- Send comment notification emails
- Admin dashboards for active users and projects
- Project and workspace invites with notifications
- Implicit saving for chains
- Improved test row selector
- Text-based diffing in Compare tool for prompts, chains and endpoints
- Support for GPT-3.5 16k and GPT-4 32k
- Analytics dashboards for endpoint requests
- Support for fine-tuned models (GPT-3.5)
- Support for vector store queries (Pinecone)
- Improved model selector
- Edit/Test mode for chains
- Compare tool
- Google Analytics
- Improved chain UI
- Support for Functions
- Support for System prompt
- UI for chain versions
- Waitlist flow
- Improved API key flow
- Versioning for chains
- Logs for endpoint requests
- Initial version ready to demo