AI Private Workspace v0.2.3

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

A change-aware release. The app now keeps track of what changed in your
project and keeps the local model's knowledge fresh — without re-indexing the
whole repository. A dated change history (git-based, cheap), an incremental
re-index of only the files whose content changed, and an Ask that can answer
"what changed today / since yesterday" from that journal. The bundled llama.cpp
engine is also faster and more reliable, and the read-only Investigator agent
uses schema-constrained output where the engine supports it. Everything stays
local, and no new model downloads are required.

Added

• Faster, more reliable bundled llama.cpp engine. The answer server now starts with Flash Attention (quicker generation, smaller KV cache) and a KV-slot save path so a warm prompt prefix can persist to disk. Both are best-effort: if a particular build rejects a flag, the server automatically retries on the known-good baseline, so startup can never break. New env knobs: AI_WORKSPACE_LLAMA_FLASH_ATTN (default on) and AI_WORKSPACE_LLAMA_PARALLEL (default 1; raise for concurrent request slots, at the cost of per-request context). GPU offload was deliberately left out to stay safe on the CPU-only Windows/Linux builds.
• Structured (JSON-Schema) output for the llama.cpp provider. Generation can be constrained to a JSON Schema so the model cannot emit invalid JSON (core.domain.structured_output) — the foundation for more reliable agents and structured features. (Ollama accepts the option but ignores it for now.)
• Exact token budgeting. When the engine exposes a tokenizer (llama.cpp /tokenize), the Ask context-window budget now uses real token counts instead of the ~4-chars-per-token estimate, and falls back to the estimate otherwise.
• More reliable Investigator steps. When the engine supports structured output (llama.cpp), the read-only Investigator agent now asks for each ReAct step as a schema-constrained JSON object, so the chosen tool-call can't be malformed. Engines without it (Ollama) keep the lenient text protocol, and a non-JSON reply degrades gracefully back to text parsing — no behaviour change for them.
• Incremental re-index (changed files only). When the project changes, you can update the AI's knowledge without re-indexing the whole repo: a new "Update index (changed files)" action (Settings) re-embeds only the files whose content hash changed and drops chunks for changed/removed files, leaving the rest untouched. Tracked by a per-file content-hash manifest (no reliance on mtime). New POST /workspaces/{id}/index/changed; the vector stores gained delete-by-source-path. The Settings section shows a stale-index hint ("N files changed since the AI last indexed them", from a cheap embed-free …/index/changed/preview) and an optional auto-update (off by default) that runs the incremental update once a day when you open the workspace. Then just ask in Ask — questions like "what changed today?" or "what changed since yesterday?" now pull the dated git change journal into the answer's context automatically.
• Change history timeline. A new History tab under Project Intelligence keeps a durable, date-grouped journal of what changed in the project — summary, risk/structural counts, changed areas (hover a chip for the files) and commit subjects — so the "what changed since last time" digest no longer vanishes when you leave the view. It fills automatically: opening the app on a new day records a dated entry from git alone (a cheap path — no file rescan, no graph rebuild, no re-indexing). "Check now" additionally re-scans to capture structural changes. A generated one-tap summary is saved onto its entry too.

Changed

• New app icon. Redesigned the macOS/Windows app icon as a proper rounded-tile icon: a deep-forest gradient background with the pigeon-and-shield mark filling the tile and the home outline glowing as a frame — far more legible at dock/taskbar sizes than the previous white-background version. Regenerated the full set (.png, .ico, .icns). The brand mark also gets brighter, livelier eyes.
• Bundled llama.cpp updated to b9789 (from b9777) for scripts/fetch_llama_server.sh and the release / Windows build workflows.

All commits in this release

• release: 0.2.3
• ruff format
• feat(index): stale-index hint + optional auto-update (changed files)
• refactor(settings): extract UpdateIndexSection (CodeFactor complexity)
• ruff fix
• feat(ask): answer 'what changed today' from the dated change journal
• feat(index): incremental re-index endpoint + Settings button + DI/inventory
• feat(index): incremental re-index by content hash (backend)
• refactor(watch): single home for history — remove inline timeline from Home card
• feat(watch): cheap git-only history journal + date-grouped timeline
• refactor(watch): split history timeline into HistoryEntryItem + HighlightList
• ruff fix
• feat(watch): hover an area chip to see which files changed
• feat(watch): surface change history on the Project Watch card + richer entries
• refactor: use contextlib.suppress for best-effort history writes (B110)
• ruff fix
• ruff format fix
• docs(api): inventory the watch/history endpoint
• feat(intelligence): durable change-history timeline + History tab
• feat(investigator): JSON-schema structured steps with text fallback
• feat(llamacpp): flash-attn, KV-slot cache, parallel slots, JSON-schema output, exact tokenizer
• chore: new app icon + bump bundled llama.cpp to b9789

Full changelog: v0.2.2...v0.2.3

AI Private Workspace v0.2.2

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

A focused Ask quality release: the same local, private RAG, but the answers
are better grounded and the model's context window is used more deliberately. All
improvements are deterministic or best-effort, require no new model downloads,
and keep the default fast-start experience (the reranker stays off). The new
source-aware chunks take effect after a reindex (scan, then build).

Added

• Source-aware chunks. Every indexed chunk now carries a deterministic one-line header — [source: path › section · part N/M] — derived from its own structure (a heading or a definition name). The header is stored and shown with the chunk so the model grounds and cites more reliably and the path is keyword-searchable, while the dense embedding is computed on the clean body so similarity is unaffected. No LLM call at index time; takes effect after a reindex.
• Hallucinated-citation guard. When an answer cites a file in backticks that wasn't in the retrieved context, Ask attaches a non-blocking "verify these" note instead of silently trusting it.
• Optional query rewrite. Ask can distil a question into a compact search query with the loaded model before retrieval, so pronoun- or intent-phrased questions still find the right files. Off by default (one extra call per ask); opt in via AI_WORKSPACE_ASK_QUERY_REWRITE. Reuses the loaded model — no new download.

Changed

• Smarter use of the model's context window in Ask. The grounded prompt is now token-budgeted to the model's real context window — retrieved chunks are sized to what actually fits after reserving room for project memory, conversation history and the answer, so the engine never silently truncates an overflowing prompt. The prompt is also reordered (stable instructions/context first, the question last) and llama.cpp prompt caching is enabled, so multi-turn answers start faster.
• Semantic project memory. Recalled memory is re-ranked by embedding similarity (when a local embedder is available), so a question about "production" surfaces a note that says "prod is called prd" even with no shared words. Best-effort, with the deterministic keyword + pin + recency selection as the fallback.
• Conversation history fits the window, with rolling summarization. Recent turns are kept by token budget (not a fixed count), and older evicted turns are folded into a short running summary — so a long chat no longer forgets its earlier context after a handful of exchanges. Best-effort; falls back to just the recent turns.
• More, more diverse context per answer. With the reranker off (the default), Ask now pulls a wider candidate pool and selects a relevant-but-diverse subset with Maximal Marginal Relevance, so a fixed context budget covers more of the codebase instead of near-duplicate top hits.

All commits in this release

• release: 0.2.2
• style: ruff format test_contextual_chunk
• fix(index): embed clean chunk body, keep provenance header for display (B)
• style: ruff format (post-merge)
• style: ruff format
• docs(changelog): note answer-quality batch (chunk headers, citation guard, MMR, query rewrite)
• feat(ask): optional LLM query rewrite before retrieval (D)
• feat(index): contextual chunk headers (source path + section) (B)
• feat(ask): MMR diversification + wider candidate pool (A)
• feat(ask): flag hallucinated source citations (C)
• feat(ask): token-budgeted conversation history with rolling summarization
• docs(changelog): note Ask context-budgeting + semantic memory (Unreleased)
• feat(memory): semantic re-rank of project memory (synonyms/paraphrase)
• perf(ask): token-budget the grounded prompt to the model window + KV-cache-friendly layout
• ci(osv): apply root osv-scanner.toml via --config so triage ignores take effect
• security(deps): floor idna/Pygments; triage unfixable OSV advisories

Full changelog: v0.2.1...v0.2.2

AI Private Workspace v0.2.1

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

Added

• Build provenance & dependency scanning. Release installers now carry a cryptographic GitHub Artifact Attestation (verifiable with gh attestation verify <file> --repo …), and a weekly OSV-Scanner workflow reports known-vulnerable dependencies to Security › Code scanning. Both actions are pinned by commit SHA.

Changed

• Docs: the deep Project Intelligence / Watcher / Investigator / Memory detail moved to docs/PROJECT_INTELLIGENCE.md; the README keeps a short overview with a link. The Project Intelligence Security tab is renamed Security review.

All commits in this release

• release: 0.2.1
• ci(security): build-provenance attestations for installers + OSV-Scanner workflow
• docs+ux: slim README intelligence section into docs/PROJECT_INTELLIGENCE.md; rename Security lens -> Security review
• docs(security): refresh SECURITY.md for v0.2.x (read-only positioning, clearer out-of-scope)

Full changelog: v0.2.0...v0.2.1

AI Private Workspace v0.2.0

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

First minor release since 0.1.99. It turns AI Private Workspace from a local RAG
chat into a full read-only project-intelligence tool — still entirely local,
private, and grounded in your own files — and hardens the project's open-source
security posture. Headline themes (the detailed 0.1.100–0.1.178 history is
preserved below):

Added

• Project Intelligence. A deterministic, role-neutral evidence graph with an interactive Map; role lenses (Developer / DevOps / Tester / Business analyst) and an adaptive role dashboard; a visual CI/CD flow; Cloud and References tabs; environment comparison and deployment flow; and a read-only Security lens. Findings everywhere read as human leads ("what was found, why it may matter, what to check"), not verdicts.
• Read-only agents & analysis. The Investigator — a bounded ReAct loop over read-only tools (search_code, read_file, graph_query, list_files, git_history, ci_triggers) with a transparent trace and collected sources — and the deterministic Watcher ("what changed since I last looked?", with an optional one-tap LLM recap of the commits).
• Project groups. Treat several repositories as one project: group Ask (repo-tagged sources), a portfolio Home, and group Intelligence that compares rather than merges (repo×environment matrix, shared-vs-unique tech, risks grouped by pattern).
• Memory & profile. Local, fully-editable project memory + handbook, and a cross-project "About you" profile applied to every project — with review-first suggestion (nothing saved until you keep it).
• Navigation. A Cmd/Ctrl-K command palette and a file inspector (owner, change-coupling, blast radius, the risks touching a file).
• Answer feedback that teaches the project (👍/👎 → corrections fed back into prompts) and much richer git intelligence (activity over time, branch strategy, change coupling, merge/PR activity).
• Embedders & models. BGE-M3 / Qwen3-Embedding installable on both engines; Qwen3 4B as the default answer model; conversation memory and an opt-in reranker in Ask.

Security & supply chain

• OpenSSF Best Practices: passing, plus OpenSSF Scorecard, CodeQL SAST, REUSE-compliant licensing, and a CodeFactor grade. All GitHub Actions are pinned to commit SHAs; workflow tokens follow least privilege; the backend Docker image is pinned by digest. Releases are signed and ship SHA256SUMS, an SPDX SBOM, and an automated-test report. Contact routes through GitHub (no email in the repo).

Changed

• The Skills library is now a single editor in Settings; the Models tab gains Tuning. Bundled llama.cpp updated to b9777.

All commits in this release

• release: 0.2.0
• chore(release): 0.1.178 — OpenSSF passing badge, Scorecard hardening, build fixes
• fix(lint): resolve GitChangeBrief forward-ref annotations under TYPE_CHECKING (ruff F821)
• fix(build): vite manualChunks as function form (type-compatible after Vite/Rollup bump)
• docs(readme): add OpenSSF Best Practices passing badge (project 13357)
• chore(deps): update httpx requirement in /backend
• chore(deps): update pathspec requirement in /backend
• chore(ci): bump actions/checkout from 6.0.2 to 7.0.0
• build: bump backend Docker base image to python:3.14-slim (matches local), pinned by digest
• chore(deps): bump the frontend-dev group in /frontend with 4 updates
• chore(deps): bump time from 0.3.36 to 0.3.51 in /frontend/src-tauri
• chore(ci): bump actions/upload-artifact from 4.6.2 to 7.0.1
• chore(deps): update pydantic requirement in /backend
• chore(ci): bump actions/setup-python from 6.2.0 to 6.3.0
• chore(ci): bump github/codeql-action from 3.36.0 to 4.36.2
• chore(deps): update qdrant-client requirement in /backend
• chore(deps): update fastapi requirement in /backend
• chore(deps): bump python from 3.11-slim to 3.14-slim in /backend
• build(security): pin Docker base image by digest; add docker ecosystem to Dependabot
• ci(security): least-privilege token perms in release.yml; link content in SECURITY.md
• docs(readme): move CI to top row, license to the compliance row
• docs(readme): use native REUSE + CodeFactor badges (shields proxy showed stale/unregistered)
• docs(readme): group badges into two centered rows, unify badge style
• chore: route contact through GitHub, drop email from copyright
• chore(security): set up OpenSSF/REUSE/CodeQL badges + pin actions; bump llama.cpp b9777
• feat(watch): one-tap LLM recap of commits since last check
• feat(watch): git-grounded human change brief in 'What changed since last time'
• fix(watch): correct plural in change digest (dependencys -> dependencies)
• fix(routes): import get_settings in workspaces.py (NameError in understanding route)
• feat(analysis): fit project-understanding prompt to the context window
• fix(llamacpp): raise answer-model context to 8192 (was 4096)
• fix(index): tag untagged index on embedder switch so reindex prompt is reliable
• fix(index): show reindex prompt only when the search model changed
• refactor(skills): single editor in Settings; remove Models skill duplicate
• fix(models): embedding provider reports the engine's actual model on startup/open
• chore(ui): About-you Style placeholder example -> 'Answer in Ukrainian'. 0.1.166
• feat(brand): use light/dark pigeon icons for tray + in-app logo, theme-swapped
• fix(models): re-point live embedding provider when switching search model
• feat(desktop): system tray icon (mac menu bar + windows), theme-aware on macOS
• feat(index): structure-aware chunking (Python AST + brace-aware + fenced markdown)
• fix(setup): don't bounce to onboarding takeover when switching the search model
• fix(ui): theme-aware in-app logo (dark logo in dark mode, light in light)
• fix(models): refresh 'current setup' after switching llama.cpp answer/search model
• fix(models): switch + delete the built-in search/embedding model
• fix(models): re-attach to running llama.cpp download after panel remount
• test: guided setup default is now qwen3:4b
• feat(models): make new embedders installable on both engines + refresh default to Qwen3 4B
• feat(memory): review-first profile-fact suggestions
• chore(models): drop mxbai embedder suggestion + fix catalog count tests
• ui(models): clearer at-a-glance embedder comparison + hover tip
• feat(memory): user profile — cross-project memory about the person
• feat(embeddings): add BGE-M3 + Qwen3-Embedding 0.6B to the catalog
• ui(ask): remove the Assistant focus panel
• ui(settings): open skills editor on the workspace's active role
• fix(roles): role change also sets the active Ask skill
• feat(roles): unify to 5 canonical roles + persist role as one workspace setting
• feat(role): land on role's tab + mark 'for your role' risks
• fix(role-brief): de-duplicate top risks by title in the dashboard band
• fix(groups): drop deleted workspaces from group member counts
• ui: drop duplicate risk badge, de-dupe Overview counts, fix header bleed
• ui: calmer risk cards + design-token normalisation (Jony pass)
• docs: add repo health audit report
• docs: actualize README (main-vs-release note, test suite + CI report); changelog
• ci: render backend test report as a job summary + JUnit artifact
• chore(frontend): remove 52 unused doc-mirror TS types
• feat(env+docs): environment matrix table; read-only-first README
• feat(cicd): visual CI/CD flow tab (trigger -> workflows -> jobs)
• feat(dashboard): adaptive role dashboard brief + clickable suggested questions
• feat(role): role brief + evidence-gated suggested questions + role-shaped Ask
• feat(risks): human-readable, review-oriented risk explanations
• docs: reframe README as read-only project intelligence; remove MCP from public copy
• docs: fully remove tile-outline remnants from pigeon mark (firm top/side sweep)
• docs: clean pigeon mark — remove black border stroke + white tile, keep feet
• docs: bigger transparent pigeon mark at README top (white tile + halo removed)
• docs: greet visitors with the pigeon logo at the top of the README
• 0.1.133: docs — refresh install + welcome screenshots with new pigeon logo
• chore: sync Cargo.lock to 0.1.132
• 0.1.132: docs — refreshed, redacted README screenshots; gitignore raw screenshot drop
• 0.1.131: pigeon avatar in Ask + caret toggle for sources (drop 'Show')
• 0.1.130: fix empty groups on launch + duplicate group on drag (load after backend ready, fresh-list dedup)
• 0.1.129: declutter Ask — quiet 'Answer style' link, smaller sources bar
• 0.1.128: consistency pass — unified eyebrows/chips/cards, single accent token, calmer Ask empty state
• 0.1.127: calmer Home — clear lead line + Ask path, secondary cards collapsed by default
• docs: refresh screenshot capture guide for pigeon icon + new tabs (Security/Map, file inspector, command palette, groups)
• 0.1.126: docs — refresh README with groups, command palette, file inspector, coupling, security lens
• 0.1.125: clickable files in Intelligence 'where to start' + honest empty state for sparse projects
• 0.1.124: label the handbook clearly (auto-fed into answers + 'In use' badge)
• 0.1.123: make file inspector discoverable (clickable start-here + hotspots) + clarify role-applies-instantly
• 0.1.122: security posture lens (CI scanners + security-relevant findings)
• 0.1.121: file inspector (ownership, coupling, blast radius, risks) + palette file search
• 0.1.120: surface change ...

AI Private Workspace v0.1.169

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

All commits in this release

• fix(index): show reindex prompt only when the search model changed
• refactor(skills): single editor in Settings; remove Models skill duplicate
• fix(models): embedding provider reports the engine's actual model on startup/open
• chore(ui): About-you Style placeholder example -> 'Answer in Ukrainian'. 0.1.166
• feat(brand): use light/dark pigeon icons for tray + in-app logo, theme-swapped
• fix(models): re-point live embedding provider when switching search model
• feat(desktop): system tray icon (mac menu bar + windows), theme-aware on macOS
• feat(index): structure-aware chunking (Python AST + brace-aware + fenced markdown)
• fix(setup): don't bounce to onboarding takeover when switching the search model
• fix(ui): theme-aware in-app logo (dark logo in dark mode, light in light)
• fix(models): refresh 'current setup' after switching llama.cpp answer/search model
• fix(models): switch + delete the built-in search/embedding model
• fix(models): re-attach to running llama.cpp download after panel remount
• test: guided setup default is now qwen3:4b
• feat(models): make new embedders installable on both engines + refresh default to Qwen3 4B
• feat(memory): review-first profile-fact suggestions
• chore(models): drop mxbai embedder suggestion + fix catalog count tests
• ui(models): clearer at-a-glance embedder comparison + hover tip
• feat(memory): user profile — cross-project memory about the person
• feat(embeddings): add BGE-M3 + Qwen3-Embedding 0.6B to the catalog

Full changelog: v0.1.150...v0.1.169

AI Private Workspace v0.1.150

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

All commits in this release

• ui(ask): remove the Assistant focus panel
• ui(settings): open skills editor on the workspace's active role
• fix(roles): role change also sets the active Ask skill

Full changelog: v0.1.148...v0.1.150

AI Private Workspace v0.1.148

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

All commits in this release

• feat(roles): unify to 5 canonical roles + persist role as one workspace setting
• feat(role): land on role's tab + mark 'for your role' risks
• fix(role-brief): de-duplicate top risks by title in the dashboard band

Full changelog: v0.1.145...v0.1.148

AI Private Workspace v0.1.145

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

All commits in this release

• fix(groups): drop deleted workspaces from group member counts
• ui: drop duplicate risk badge, de-dupe Overview counts, fix header bleed

Full changelog: v0.1.143...v0.1.145

AI Private Workspace v0.1.143

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

All commits in this release

• ui: calmer risk cards + design-token normalisation (Jony pass)
• docs: add repo health audit report
• docs: actualize README (main-vs-release note, test suite + CI report); changelog
• ci: render backend test report as a job summary + JUnit artifact
• chore(frontend): remove 52 unused doc-mirror TS types
• feat(env+docs): environment matrix table; read-only-first README
• feat(cicd): visual CI/CD flow tab (trigger -> workflows -> jobs)
• feat(dashboard): adaptive role dashboard brief + clickable suggested questions
• feat(role): role brief + evidence-gated suggested questions + role-shaped Ask
• feat(risks): human-readable, review-oriented risk explanations
• docs: reframe README as read-only project intelligence; remove MCP from public copy
• docs: fully remove tile-outline remnants from pigeon mark (firm top/side sweep)
• docs: clean pigeon mark — remove black border stroke + white tile, keep feet
• docs: bigger transparent pigeon mark at README top (white tile + halo removed)
• docs: greet visitors with the pigeon logo at the top of the README
• 0.1.133: docs — refresh install + welcome screenshots with new pigeon logo
• chore: sync Cargo.lock to 0.1.132
• 0.1.132: docs — refreshed, redacted README screenshots; gitignore raw screenshot drop
• 0.1.131: pigeon avatar in Ask + caret toggle for sources (drop 'Show')
• 0.1.130: fix empty groups on launch + duplicate group on drag (load after backend ready, fresh-list dedup)
• 0.1.129: declutter Ask — quiet 'Answer style' link, smaller sources bar
• 0.1.128: consistency pass — unified eyebrows/chips/cards, single accent token, calmer Ask empty state
• 0.1.127: calmer Home — clear lead line + Ask path, secondary cards collapsed by default
• docs: refresh screenshot capture guide for pigeon icon + new tabs (Security/Map, file inspector, command palette, groups)

Full changelog: v0.1.126...v0.1.143

AI Private Workspace v0.1.126

Downloads

macOS

• Apple Silicon (arm64)
• Intel (x64)

Windows

• Windows x64 installer (.exe)

Documentation

• README refreshed to cover the capabilities added recently: multi-repo project groups (portfolio Ask/Home/Intelligence), the Cmd/Ctrl-K command palette, the file inspector, the read-from-git project activity card with change coupling, and the security lens.

All commits in this release

• 0.1.126: docs — refresh README with groups, command palette, file inspector, coupling, security lens
• 0.1.125: clickable files in Intelligence 'where to start' + honest empty state for sparse projects
• 0.1.124: label the handbook clearly (auto-fed into answers + 'In use' badge)
• 0.1.123: make file inspector discoverable (clickable start-here + hotspots) + clarify role-applies-instantly
• 0.1.122: security posture lens (CI scanners + security-relevant findings)
• 0.1.121: file inspector (ownership, coupling, blast radius, risks) + palette file search
• 0.1.120: surface change counts in 'what changed since last time' digest
• 0.1.119: command palette (Cmd/Ctrl-K) for repos, groups, sections
• 0.1.118: group view full-width + default branch in group repo cards
• 0.1.117: render markdown in group Ask answers
• 0.1.116: group Intelligence compares (matrix/shared-vs-unique/grouped risks) + Ask polish
• 0.1.115: drag-to-group — auto rename on create, dedup, no-op on existing membership
• 0.1.114: change coupling + role-aware git activity card
• Remove 'When the work happens' weekday rhythm (low signal); keep branch-flow facts
• 0.1.113: richer git activity card — weekday rhythm + branch flow
• 0.1.112: Map blast radius, group fixes & DnD, icon Manage menu, editable memory
• Rating nudges: log 👍/👎 + deterministic suggestions (bigger model / rebuild context) in Ask

Full changelog: v0.1.111...v0.1.126