A local-first, multi-provider Streamlit AI workbench for local, remote and cloud LLMs.
- Candidate-blocked duplicate analysis
- Paginated Knowledge documents
- Safe canonical selection and recoverable duplicate archive
- Operational Knowledge workspace
- Exact and near-duplicate grouping
- Source metadata
- Batch stale/new catalog refresh
- Index health and workflow logging
- Token adapter interface
- Attachment size/truncation diagnostics
- Improved request preview and utilisation health
- Automatic memory thresholds
- Smarter context selection
- Configurable recent-turn retention
- Optional request preview
- Inline provider/model readiness
- Inline Knowledge toggle
- Universal context surface
- Three-tab Chat layout
- Responsive/touch hardening
- Live Research Intelligence inspector
- Conversation-first home screen
- Minimal sidebar
- Progressive disclosure
- Central Apps & Tools hub
- Persistent Knowledge catalog
- SHA-256 content fingerprints
- Indexed/stale lifecycle visibility
- Workspace Knowledge statistics
- Exact duplicate-content detection
- Searchable catalog and refresh workflow
Multi-provider chat, true streaming, Workspaces, Prompt Library, Knowledge retrieval, semantic/hybrid retrieval, Research, Model Compare, Model Lab, Structured Output, versioned Artifacts, Usage, resilient routing, backup/restore, privacy, attachments, conversation memory and Help Centre.
On Windows, double-click run_windows.bat, or run streamlit run app.py. Data remains in %USERPROFILE%\.peabox_streamlit_llm and credentials remain outside release archives.
See USER_GUIDE.md, FEATURES.md, CHANGELOG.md, VERSION_HISTORY.md, ROADMAP.md, RELEASE_NOTES_v7.12.0.md, and help/.