Repository navigation
Releases
Releases · yottadynamics/yottacode
v0.5.0
Changelog
- abd80d9 Add ACP model and session controls (#388)
- 7ac2fcc Add GA document tools and Python helpers (#225)
- f62921a Add JSON receipts for one-shot runs (#271)
- e2466db Add JSONL session activity exports (#266)
- d506215 Add OAuth 2.1 support for remote MCP servers Wires the OAuth 2.1 authorization-code + PKCE client already vendored in the pinned modelcontextprotocol/go-sdk (RFC 8707 resource indicators, RFC 9728 protected-resource discovery, refresh) into HTTPClient, gated behind a new auth = oauth server config. Adds /mcp auth|logout and yottacode mcp auth|logout, persists tokens to ~/.yottacode/mcp-auth/.json, and keeps the live session transport non-interactive (a missing/expired/revoked token fails fast with guidance instead of popping a browser during unattended background startup or a mid-turn tool call) — only the explicit /mcp auth command runs the interactive browser flow. Verified against Google's live Gmail MCP protected-resource-metadata endpoint; targets Google's Workspace MCP servers as the first real consumer. (#329)
- 54560ae Add PDF OCR and document search (#242)
- a0b8fcb Add TUI terminal title status (#277)
- e24cd04 Add build-tagged deadlock detection (#348)
- 0c464d3 Add conservative worktree cleanup command (#344)
- bfd5fba Add default commit and PR attribution (#315)
- 2496b1c Add deterministic deep research workflow (#382)
- f275a43 Add experimental browser automation tools (#320)
- be5347e Add explicit Code Map context assembly (#319)
- 6323a85 Add fine-grained permission grants (#283)
- 674b5c0 Add hashline edit reliability (#280)
- 8480498 Add headless memory recall command (#311)
- eaf37d9 Add interactive session inspect view (#387)
- 53f07ec Add models.dev usage cost estimates (#325)
- 845f980 Add official sandbox image (#253)
- 6cff55a Add pptx table extraction (#241)
- e99e5ca Add provider headers and OpenRouter setup preset (#314)
- 0b6d3c5 Add provider setup issue template (#269)
- 20c34e4 Add shell job transcript logs (#397)
- 2c65093 Add structured JSON output for oneshot runs (#317)
- fe4a468 Add system permissions policy (#349)
- 5f25160 Add worktree exit cleanup prompt (#267)
- aae068f Align TUI doctor with CLI diagnostics (#369)
- c88f4cc Avoid real podman probes in sandbox picker test (#229)
- 5504908 Block interactive privilege prompts in run_bash (#300)
- a080350 Bound TUI live preview rendering work (#347)
- 9f3d710 Bound ffmpeg resource usage for media tools (#296)
- 11ffd58 Coalesce foreground subagent progress in TUI (#365)
- 4332f8f Compact approval cards on wide terminals (#368)
- ddfa632 Default sandbox networking to host (#268)
- 20bc9b7 Dim cmdline cursor while agent is working (#361)
- f40f94a Enhance document image metadata extraction (#273)
- 27e8860 Enrich session list stats (#276)
- 6d7ab79 Expand browser automation capabilities (#352)
- 45c7d06 Fix ACP cancel test start signal flake (#262)
- 770105c Fix PR Markdown formatting in TUI (#334)
- 68112c8 Fix PR update formatting and approval spacing (#330)
- e1da4e5 Fix TUI approval and popup state cleanup (#379)
- bdcef50 Fix approval modal bash layout (#383)
- 2173eb5 Fix clipped approval modal previews (#250)
- e386a6a Fix clipped approval modal previews (#275)
- 9734c33 Fix context compaction recovery and exit timeout (#265)
- f9c91f6 Fix core edit integrity and tool reliability (#390)
- d4adde9 Fix custom endpoint key field and clip validation glyph to width (#309)
- 721c8d5 Fix duplicate approval modal title (#264)
- d6e064f Fix git push for mismatched worktree upstreams (#355)
- 7718810 Fix lazy sandbox profile recovery (#238)
- 915792f Fix memory browser pagination and popup sizing (#378)
- 1fb04c3 Fix openai-auth 401s from missing chatgpt-account-id header (#371)
- 80023c5 Fix pasting into TUI popups after bubbletea v2 migration (#306)
- 38748a3 Fix raw GitHub URL skill installs and add Custom catalog tab Private-repo raw.githubusercontent.com URLs carry a ?token=... query string that only authorizes that one blob fetch, but install routed them through the codeload.github.com archive path, which discards the token and 404s. Route query-bearing raw URLs through the single-file fetch instead. Also: CLI installs of official/local skills now persist into config.toml's [skills] default_on so they're enabled next session without a separate /skills trip, matching the in-TUI install path. Adds a Custom tab to /skills > Catalog for skills installed from anything other than the official catalog (previously invisible there), plus an -bound local rescan on Bundled/Custom so a CLI install shows up in an already-open session without restarting. Claude-Session: https://claude.ai/code/session_01VmadWnhkUyMigu6ff4J7Pb (#326)
- 8626dc2 Fix sandbox DNS and Go test temp dirs (#270)
- 09f2e66 Fix stale router pair lockout via graceful degradation (#307)
- de98760 Fix startup update check blocking TUI (#285)
- a46b94e Fix welcome hover after clear (#223)
- a7d87b9 Graduate command sandbox to GA (#240)
- acaf255 Harden ACP session, memory, and context handling (#384)
- 8b6880c Harden MCP transports and lifecycle management (#321)
- 6b8aea6 Harden agent loop and subagent lifecycle against races and leaks (#327)
- b44a272 Harden background subagent permissions (#336)
- 0f8f583 Harden browser automation and add human handoff (#356)
- e9c5c65 Harden browser stealth and tab setup (#374)
- 492a3bb Harden compaction and concurrent mutations (#340)
- fe60be0 Harden edit tool recovery and validation (#389)
- 9d0f31f Harden edit tools and native web search (#396)
- 80c2b83 Harden file reads and anchored edits (#362)
- 646df88 Harden nested Bash permission matching (#350)
- 8912857 Harden offline syntax range selection (#316)
- 431414e Harden recall search query sanitization (#398)
- cefae41 Harden receipt-based edit reliability (#394)
- 0324926 Harden run_bash timeouts and background jobs (#395)
- 7e34c3d Harden sandbox credential secret lifecycle (#364)
- 4b00cb1 Harden sandbox image release workflow (#259)
- db98138 Harden structured edit and patch workflows (#342)
- 17ce875 Harden subagent dispatch safety (#274)
- b03104a Harden subagent resource cleanup (#323)
- 0d35ee1 Harden subprocess cancellation and blocking waits (#346)
- e53bec5 Improve advisor routing status and setup (#335)
- 470661c Improve approval command layout (#400)
- a00e324 Improve doctor output summary (#301)
- 5bd6aee Improve pull request result cards (#343)
- be16e6b Improve subagent code search efficiency (#318)
- bf1735e Instrument /inspect session turn diagnostics (#338)
- 835f0cd Isolate polluted host shell environments (#297)
- 1c139ed Keep host Go test artifacts out of repo (#288)
- 5479441 Limit TUI mouse handling to transcript scrolling (#237)
- 5f03d87 Make approval modals keyboard-only (#236)
- 2a3f8fd Make media rendering safer for desktops (#299)
- a15aacc Migrate browser automation from Rod to chromedp (#360)...
v0.5.0-rc.2
Changelog
- 7ac2fcc Add GA document tools and Python helpers (#225)
- f62921a Add JSON receipts for one-shot runs (#271)
- e2466db Add JSONL session activity exports (#266)
- d506215 Add OAuth 2.1 support for remote MCP servers Wires the OAuth 2.1 authorization-code + PKCE client already vendored in the pinned modelcontextprotocol/go-sdk (RFC 8707 resource indicators, RFC 9728 protected-resource discovery, refresh) into HTTPClient, gated behind a new auth = oauth server config. Adds /mcp auth|logout and yottacode mcp auth|logout, persists tokens to ~/.yottacode/mcp-auth/.json, and keeps the live session transport non-interactive (a missing/expired/revoked token fails fast with guidance instead of popping a browser during unattended background startup or a mid-turn tool call) — only the explicit /mcp auth command runs the interactive browser flow. Verified against Google's live Gmail MCP protected-resource-metadata endpoint; targets Google's Workspace MCP servers as the first real consumer. (#329)
- 54560ae Add PDF OCR and document search (#242)
- a0b8fcb Add TUI terminal title status (#277)
- e24cd04 Add build-tagged deadlock detection (#348)
- 0c464d3 Add conservative worktree cleanup command (#344)
- bfd5fba Add default commit and PR attribution (#315)
- f275a43 Add experimental browser automation tools (#320)
- be5347e Add explicit Code Map context assembly (#319)
- 6323a85 Add fine-grained permission grants (#283)
- 674b5c0 Add hashline edit reliability (#280)
- 8480498 Add headless memory recall command (#311)
- 53f07ec Add models.dev usage cost estimates (#325)
- 845f980 Add official sandbox image (#253)
- 6cff55a Add pptx table extraction (#241)
- e99e5ca Add provider headers and OpenRouter setup preset (#314)
- 0b6d3c5 Add provider setup issue template (#269)
- 2c65093 Add structured JSON output for oneshot runs (#317)
- 5f25160 Add worktree exit cleanup prompt (#267)
- c88f4cc Avoid real podman probes in sandbox picker test (#229)
- 5504908 Block interactive privilege prompts in run_bash (#300)
- a080350 Bound TUI live preview rendering work (#347)
- 9f3d710 Bound ffmpeg resource usage for media tools (#296)
- ddfa632 Default sandbox networking to host (#268)
- f40f94a Enhance document image metadata extraction (#273)
- 27e8860 Enrich session list stats (#276)
- 6d7ab79 Expand browser automation capabilities (#352)
- 45c7d06 Fix ACP cancel test start signal flake (#262)
- 770105c Fix PR Markdown formatting in TUI (#334)
- 68112c8 Fix PR update formatting and approval spacing (#330)
- 2173eb5 Fix clipped approval modal previews (#250)
- e386a6a Fix clipped approval modal previews (#275)
- 9734c33 Fix context compaction recovery and exit timeout (#265)
- d4adde9 Fix custom endpoint key field and clip validation glyph to width (#309)
- 721c8d5 Fix duplicate approval modal title (#264)
- d6e064f Fix git push for mismatched worktree upstreams (#355)
- 7718810 Fix lazy sandbox profile recovery (#238)
- 80023c5 Fix pasting into TUI popups after bubbletea v2 migration (#306)
- 38748a3 Fix raw GitHub URL skill installs and add Custom catalog tab Private-repo raw.githubusercontent.com URLs carry a ?token=... query string that only authorizes that one blob fetch, but install routed them through the codeload.github.com archive path, which discards the token and 404s. Route query-bearing raw URLs through the single-file fetch instead. Also: CLI installs of official/local skills now persist into config.toml's [skills] default_on so they're enabled next session without a separate /skills trip, matching the in-TUI install path. Adds a Custom tab to /skills > Catalog for skills installed from anything other than the official catalog (previously invisible there), plus an -bound local rescan on Bundled/Custom so a CLI install shows up in an already-open session without restarting. Claude-Session: https://claude.ai/code/session_01VmadWnhkUyMigu6ff4J7Pb (#326)
- 8626dc2 Fix sandbox DNS and Go test temp dirs (#270)
- 09f2e66 Fix stale router pair lockout via graceful degradation (#307)
- de98760 Fix startup update check blocking TUI (#285)
- a46b94e Fix welcome hover after clear (#223)
- a7d87b9 Graduate command sandbox to GA (#240)
- 8b6880c Harden MCP transports and lifecycle management (#321)
- 6b8aea6 Harden agent loop and subagent lifecycle against races and leaks (#327)
- b44a272 Harden background subagent permissions (#336)
- 0f8f583 Harden browser automation and add human handoff (#356)
- 492a3bb Harden compaction and concurrent mutations (#340)
- 8912857 Harden offline syntax range selection (#316)
- 4b00cb1 Harden sandbox image release workflow (#259)
- db98138 Harden structured edit and patch workflows (#342)
- 17ce875 Harden subagent dispatch safety (#274)
- b03104a Harden subagent resource cleanup (#323)
- 0d35ee1 Harden subprocess cancellation and blocking waits (#346)
- e53bec5 Improve advisor routing status and setup (#335)
- a00e324 Improve doctor output summary (#301)
- 5bd6aee Improve pull request result cards (#343)
- be16e6b Improve subagent code search efficiency (#318)
- bf1735e Instrument /inspect session turn diagnostics (#338)
- 835f0cd Isolate polluted host shell environments (#297)
- 1c139ed Keep host Go test artifacts out of repo (#288)
- 5479441 Limit TUI mouse handling to transcript scrolling (#237)
- 5f03d87 Make approval modals keyboard-only (#236)
- 2a3f8fd Make media rendering safer for desktops (#299)
- 80fbd95 Move plan approval labels inside dialog (#341)
- e92cc95 Persist sandbox Go caches across sessions (#279)
- a9b1415 Polish Ollama memory setup flow (#345)
- 3b7e202 Polish approval modal layout (#339)
- 3a19989 Polish first-contributor guide (#227)
- 3a449f7 Polish permission warnings and sandbox limits (#272)
- bd7f807 Polish setup wizard review actions (#222)
- 09f0bb6 Prepare v0.5.0-rc.1 release (#353)
- bf7d6fc Prepare v0.5.0-rc.2 release
- 18da3f9 Prevent inherited worktree upstreams (#337)
- 5407e3f Prune abandoned running sandbox containers (#286)
- d4918b4 Quiet expected dispatch panic test output (#298)
- 8d7aed5 Quiet expected panic recovery tests (#304)
- 17a765e Reduce tool context growth (#324)
- e168138 Refine git issue implementation gate (#261)
- 1cecbd4 Refresh README for launch positioning (#303)
- 9195bf7 Refresh README hero and quickstart flow (#239)
- b03ec53 Refresh indirect Go dependencies (#263)
- af146f8 Reload permissions before showing picker (#284)
- 5567722 Remove loop exit close glyph (#249)
- 25b46b0 Repair orphaned tool calls before every request; fix a git-editor hang (#328)
- 17d9e1e Restore copyable OpenAI auth URLs (#313)
- f299625 Revert clear hero reset to avoid mouse lag (#234)
- f8cb88f Route sandbox commands by tool profile (#228)
- 7bbc2be Schedule weekly vulnerability scan (#354)
- 591515b Silence expected recall validation usage (#331)
- 5280142 Skip redundant mouse motion updates (#...
v0.5.0-rc.1
Changelog
- 7ac2fcc Add GA document tools and Python helpers (#225)
- f62921a Add JSON receipts for one-shot runs (#271)
- e2466db Add JSONL session activity exports (#266)
- d506215 Add OAuth 2.1 support for remote MCP servers Wires the OAuth 2.1 authorization-code + PKCE client already vendored in the pinned modelcontextprotocol/go-sdk (RFC 8707 resource indicators, RFC 9728 protected-resource discovery, refresh) into HTTPClient, gated behind a new auth = oauth server config. Adds /mcp auth|logout and yottacode mcp auth|logout, persists tokens to ~/.yottacode/mcp-auth/.json, and keeps the live session transport non-interactive (a missing/expired/revoked token fails fast with guidance instead of popping a browser during unattended background startup or a mid-turn tool call) — only the explicit /mcp auth command runs the interactive browser flow. Verified against Google's live Gmail MCP protected-resource-metadata endpoint; targets Google's Workspace MCP servers as the first real consumer. (#329)
- 54560ae Add PDF OCR and document search (#242)
- a0b8fcb Add TUI terminal title status (#277)
- e24cd04 Add build-tagged deadlock detection (#348)
- 0c464d3 Add conservative worktree cleanup command (#344)
- bfd5fba Add default commit and PR attribution (#315)
- f275a43 Add experimental browser automation tools (#320)
- be5347e Add explicit Code Map context assembly (#319)
- 6323a85 Add fine-grained permission grants (#283)
- 674b5c0 Add hashline edit reliability (#280)
- 8480498 Add headless memory recall command (#311)
- 53f07ec Add models.dev usage cost estimates (#325)
- 845f980 Add official sandbox image (#253)
- 6cff55a Add pptx table extraction (#241)
- e99e5ca Add provider headers and OpenRouter setup preset (#314)
- 0b6d3c5 Add provider setup issue template (#269)
- 2c65093 Add structured JSON output for oneshot runs (#317)
- 5f25160 Add worktree exit cleanup prompt (#267)
- c88f4cc Avoid real podman probes in sandbox picker test (#229)
- 5504908 Block interactive privilege prompts in run_bash (#300)
- a080350 Bound TUI live preview rendering work (#347)
- 9f3d710 Bound ffmpeg resource usage for media tools (#296)
- ddfa632 Default sandbox networking to host (#268)
- f40f94a Enhance document image metadata extraction (#273)
- 27e8860 Enrich session list stats (#276)
- 45c7d06 Fix ACP cancel test start signal flake (#262)
- 770105c Fix PR Markdown formatting in TUI (#334)
- 68112c8 Fix PR update formatting and approval spacing (#330)
- 2173eb5 Fix clipped approval modal previews (#250)
- e386a6a Fix clipped approval modal previews (#275)
- 9734c33 Fix context compaction recovery and exit timeout (#265)
- d4adde9 Fix custom endpoint key field and clip validation glyph to width (#309)
- 721c8d5 Fix duplicate approval modal title (#264)
- 7718810 Fix lazy sandbox profile recovery (#238)
- 80023c5 Fix pasting into TUI popups after bubbletea v2 migration (#306)
- 38748a3 Fix raw GitHub URL skill installs and add Custom catalog tab Private-repo raw.githubusercontent.com URLs carry a ?token=... query string that only authorizes that one blob fetch, but install routed them through the codeload.github.com archive path, which discards the token and 404s. Route query-bearing raw URLs through the single-file fetch instead. Also: CLI installs of official/local skills now persist into config.toml's [skills] default_on so they're enabled next session without a separate /skills trip, matching the in-TUI install path. Adds a Custom tab to /skills > Catalog for skills installed from anything other than the official catalog (previously invisible there), plus an -bound local rescan on Bundled/Custom so a CLI install shows up in an already-open session without restarting. Claude-Session: https://claude.ai/code/session_01VmadWnhkUyMigu6ff4J7Pb (#326)
- 8626dc2 Fix sandbox DNS and Go test temp dirs (#270)
- 09f2e66 Fix stale router pair lockout via graceful degradation (#307)
- de98760 Fix startup update check blocking TUI (#285)
- a46b94e Fix welcome hover after clear (#223)
- a7d87b9 Graduate command sandbox to GA (#240)
- 8b6880c Harden MCP transports and lifecycle management (#321)
- 6b8aea6 Harden agent loop and subagent lifecycle against races and leaks (#327)
- b44a272 Harden background subagent permissions (#336)
- 492a3bb Harden compaction and concurrent mutations (#340)
- 8912857 Harden offline syntax range selection (#316)
- 4b00cb1 Harden sandbox image release workflow (#259)
- db98138 Harden structured edit and patch workflows (#342)
- 17ce875 Harden subagent dispatch safety (#274)
- b03104a Harden subagent resource cleanup (#323)
- 0d35ee1 Harden subprocess cancellation and blocking waits (#346)
- e53bec5 Improve advisor routing status and setup (#335)
- a00e324 Improve doctor output summary (#301)
- 5bd6aee Improve pull request result cards (#343)
- be16e6b Improve subagent code search efficiency (#318)
- bf1735e Instrument /inspect session turn diagnostics (#338)
- 835f0cd Isolate polluted host shell environments (#297)
- 1c139ed Keep host Go test artifacts out of repo (#288)
- 5479441 Limit TUI mouse handling to transcript scrolling (#237)
- 5f03d87 Make approval modals keyboard-only (#236)
- 2a3f8fd Make media rendering safer for desktops (#299)
- 80fbd95 Move plan approval labels inside dialog (#341)
- e92cc95 Persist sandbox Go caches across sessions (#279)
- a9b1415 Polish Ollama memory setup flow (#345)
- 3b7e202 Polish approval modal layout (#339)
- 3a19989 Polish first-contributor guide (#227)
- 3a449f7 Polish permission warnings and sandbox limits (#272)
- bd7f807 Polish setup wizard review actions (#222)
- 9077950 Prepare v0.5.0-rc.1 release
- 18da3f9 Prevent inherited worktree upstreams (#337)
- 5407e3f Prune abandoned running sandbox containers (#286)
- d4918b4 Quiet expected dispatch panic test output (#298)
- 8d7aed5 Quiet expected panic recovery tests (#304)
- 17a765e Reduce tool context growth (#324)
- e168138 Refine git issue implementation gate (#261)
- 1cecbd4 Refresh README for launch positioning (#303)
- 9195bf7 Refresh README hero and quickstart flow (#239)
- b03ec53 Refresh indirect Go dependencies (#263)
- af146f8 Reload permissions before showing picker (#284)
- 5567722 Remove loop exit close glyph (#249)
- 25b46b0 Repair orphaned tool calls before every request; fix a git-editor hang (#328)
- 17d9e1e Restore copyable OpenAI auth URLs (#313)
- f299625 Revert clear hero reset to avoid mouse lag (#234)
- f8cb88f Route sandbox commands by tool profile (#228)
- 591515b Silence expected recall validation usage (#331)
- 5280142 Skip redundant mouse motion updates (#235)
- 21c6d41 Stop /mcp auth tests from popping a real browser (#332)
- 9012cbf Unify anchored edit permissions and diagnostics (#333)
- e3025a6 Upgrade Anthropic SDK (#256)
- 8291095 Upgrade Goldmark dependency (#260)
- 2f25754 Upgrade SQLite dependency (#255)
- b1a258a...
v0.4.0
Changelog
- 56b64a3 Add /loop recurring command with agent self-stop via loop_control tool (#104)
- 70dcdd4 Add /yolo slash command and rename "permissions bypass" to "yolo mode" (#122)
- f84aff9 Add ACP server adapter (#193)
- f677619 Add Anthropic cache TTL config and mid-session cache-reset warnings (#186)
- b58169d Add GitHub cover and stabilize dispatch test (#138)
- 8e89c89 Add Google Vertex AI provider support (#109)
- 94102c6 Add LSP code action previews (#135)
- 151c352 Add LSP document symbols and signature help (#128)
- 7fb0a51 Add LSP production-readiness capabilities (#142)
- 2f0ca2a Add PR CI workflow helper tools (#179)
- 69735e7 Add TUI session inspect and usage efficiency (#203)
- f937001 Add advisor implementer model routing (#131)
- 374b9ee Add anchored edit recovery path (#156)
- 5b30920 Add approval-gated memory curation apply (#115)
- 88d14ee Add auto slash command and full mode cycle (#196)
- 68c5ee9 Add automatic semantic session recall
- 9523600 Add closing-out directive to system prompt (#182)
- d69e833 Add community security docs (#94)
- 1283674 Add context compaction recovery (#111)
- d1220ce Add early busy-turn compaction tuning (#113)
- 416f07c Add experimental Podman command sandbox (#190)
- c0c4fcf Add experimental code map (#136)
- e89280b Add experimental document generation workflow (#209)
- 1ef3f58 Add experimental syntax_range tool (#157)
- 420209f Add fullscreen TUI transcript and popups (#194)
- 94a1d93 Add govulncheck to CI (#219)
- 5263728 Add hybrid LSP syntax fallback and impact tool (#155)
- cd58440 Add inspect session picker and export (#207)
- ea10c24 Add loop_control tool for self-stopping prose loops (#108)
- d7c5ab6 Add marketing video editing tools (#132)
- 6ba82fd Add memory archive maintenance (#117)
- a059628 Add memory audit curation planning (#114)
- d62af2d Add memory curation proposals and health (#116)
- 687a100 Add official skills catalog installs (#110)
- c91d167 Add polished video composition templates (#171)
- 4ccb2fd Add prompt-driven marketing video workflow (#170)
- ebff3f7 Add read_document agent tool with Phase A structured-file extraction (#183)
- 228dc6b Add router picker and fix large tool-call output caps (#123)
- 6c307b7 Add setup wizard pickers for Ollama models (#96)
- 9ab498b Add skills progress indicators (#124)
- 9b1684e Add studio-dark recording theme (#168)
- 76cabdc Add typed PR check watching (#178)
- 3382dbf Add yottacode dark theme (#200)
- d9a9148 Advance LSP diagnostics and semantic edit tools (#133)
- 5455a51 Allow editing queued mid-turn input (#118)
- 162d07d Bias memory capture toward durable knowledge (#103)
- 50ed1f6 Bound resize history replay (#161)
- c699ba4 Bump --max-iterations default from 50 to 100 (#112)
- 84857c6 Clarify sandbox restart state (#217)
- 8c53341 Compact anchored edit error cards (#162)
- 237e097 Embed startup header in top border (#143)
- ff9dbeb Expose advisor consult to routed implementer (#139)
- 36e521c Fix NVIDIA NIM empty completion streaming (#95)
- 8e75b85 Fix OpenAI auth retry port cleanup (#97)
- 7b90229 Fix TUI scrolling and approval chrome (#195)
- 08b4fe1 Fix govulncheck security findings (#218)
- 732fd64 Fix help popup scrolling (#214)
- 40b348d Fix live branch detection for PR tools (#134)
- 0cef748 Fix sandbox picker confirmation flow (#192)
- 863f45a Fix session resume, picker gists, archive restore, image tokens (#121)
- 7301f85 Fix static popup scrolling (#220)
- acaf0db Fix transcript scratch leak before tool calls (#137)
- 9dd699c Fix usage efficiency false positives (#205)
- 6332a85 Fix usage panel cache metrics (#202)
- 3531a7a Fix welcome and help popup keyboard behavior (#215)
- 5b9fea6 Gate prompt_cache_key to ProviderOpenAI only; fix Config.CacheKey doc (#99)
- acfabd6 Gate sandbox picker on experimental flag (#216)
- 2b7299e Graduate LSP code intelligence to GA (#177)
- db13d9e Graduate syntax_range to GA covering all four LSP languages (#181)
- 6e544a9 Harden LSP safety and smoke coverage (#150)
- 6a8a3b4 Harden approval security controls (#125)
- 9209636 Harden experimental LSP code intelligence (#126)
- 1993acc Harden experimental LSP production behavior (#169)
- 0f92b5f Harden experimental dispatch fan-out (#160)
- d2981d2 Harden memory save and agent search (#164)
- 0a5fa38 Harden memory saves and search consistency (#163)
- 4a6fcb3 Harden tool argument tolerance (#129)
- b1e1b0c Improve TUI mode indicators (#199)
- e8bad19 Improve TUI mouse selection and palette hover (#198)
- 8c8aeeb Improve approval modal transcript scrolling (#204)
- abda38e Improve popup mouse scrolling (#206)
- 64937b1 Improve recall picker UX (#107)
- e75408b Improve usage outlier attribution (#201)
- 5e13bf2 Migrate TUI to Charm v2 (#191)
- e81e8dc Page session and inspect pickers (#211)
- 80698f8 Polish LSP advisory card (#130)
- c544502 Polish README traction positioning (#174)
- 7a4476c Polish TUI card and picker rendering (#105)
- ffa8d8c Polish TUI cards and patch failure notices (#154)
- 5a87377 Polish TUI context and tool-card notices (#152)
- 969573b Polish TUI menu and routing status (#140)
- 055d8ba Polish TUI session stream readability (#159)
- fd846f2 Polish TUI status bar and usage details (#197)
- b93bcb6 Polish TUI system notices and interrupt UX (#149)
- 57c0cd8 Polish inline TUI submenu overlays (#176)
- 265cedb Polish inspect and sessions pickers (#208)
- e32c5ca Polish usage panel layout (#102)
- 1033f94 Prepare v0.4.0 release (#221)
- b9b8f24 Promote background subagents and add code review (#98)
- e72307f Promote background subagents and add local code review (#79)
- 9072e4d Prompt for LSP server installs safely (#175)
- 85613e7 Raise max-iterations default to 128 (#212)
- 2c9cc35 Redesign README for adoption (#172)
- 54f17cd Redesign TUI welcome screen (#213)
- 52b019a Refine TUI slash picker and status chrome (#148)
- 1bdc607 Reframe LSP advisory card from error to detection-first copy (#180)
- 535e291 Rename /router to /advisor in the TUI (#146)
- 53fcbe1 Render system notices as one-line rows (#151)
- 30c7a12 Replace static candidate list with live model catalog fetch for openai-auth (#189)
- 148eb2b Revert TUI polish while preserving behavior fixes (#141)
- 344f06e Serialize dispatch worktree cleanup (#210)
- 039f61c Stabilize TUI stre...
v0.3.1
Changelog
- d69e833 Add community security docs (#94)
- 6c307b7 Add setup wizard pickers for Ollama models (#96)
- 36e521c Fix NVIDIA NIM empty completion streaming (#95)
- 8e75b85 Fix OpenAI auth retry port cleanup (#97)
- 03fbb2f chore: bump version to 0.3.1
- c374572 docs: bump README version references to v0.3.0 (#92)
- 1ff5592 feat: add /git-create-issue command for creating GitHub issues (#74)
v0.3.0
Changelog
- f6f3b06 Add Git provider and more workflow tools (#17)
- fd635c7 Add GitHub integration (#35)
- 7867d45 Add MCP client (#27)
- 4337ca2 Add ability to install new skills from the TUI (#46)
- 9087c82 Add advanced memory capabilities and updates (#39)
- e2a4465 Add background subagents capabilities (#64)
- 895a6b9 Add cache-safe model routing (off by default) (#54)
- 96a5ab4 Add go routine for semantic retrieval (#73)
- 5ffe655 Add installer script (#14)
- 8291194 Add line number enhancements for read_file tool (#22)
- 3f465d5 Add live plan card and enforce todo_write design boundary (#18)
- ffc6c24 Add memory options (#45)
- 4dfae8e Add model health check (#59)
- a165b0a Add schema for data that is not well formated (#62)
- 15b29de Add skills capabilities + built-in skills (#30)
- b7eaff2 Add status bar effort and PR chips (#84)
- 4edccb0 Add theme support + themes (#33)
- d1fd221 Anthropic prompt caching: stable-head breakpoint survives per-turn memory churn (#58)
- dbdea55 Apply multiple fixes and improvements across the codebase (#65)
- 5e9c30a Context-window resolution via models.dev data (#63)
- 6cfc901 Feature/polish gemini provider (#90)
- a41397f Feature/usage command (#56)
- 5201e9e Feature/worktrees (#25)
- 9e6e93a Feature: Adjust TUI contrast for better readability (#68)
- d4e1efc Fix context window for openAI auth provider (#89)
- a9c532e Fix how skills are being calculated in /context (#61)
- 3693e97 Fix/memory bugs vector (#52)
- 9fffa58 Fix/small polishing fixes (#55)
- b4a2044 Fix/test fixes path (#34)
- 8e66a85 Git tool refresh: flag-aware read-only policy, risk-tiered approvals, six review (#88)
- 832a6e2 Harden tool errors and stream recovery (#83)
- 474e452 Memory prompt improvments (#71)
- b1e4239 Merge memory tree: unify user/project scopes under ~/.yottacode/memory/ (#75)
- 7f90159 Polish skills and fix the /context how is calculating the skills context (#60)
- 7b7390a Refine TUI and approval modal interactions (#21)
- 019f5b0 Refine install.sh and update quickstart docs (#16)
- 2aed57b Refine todo prompt gating and user echo rendering (#85)
- 0f4dd2b TUI polish: palette matching + width alignment, NO_COLOR auto-theme, intraline diff emphasis, green chevrons, dead-code sweep (#87)
- ea8501d TUI polishing changes (#44)
- 37888e6 TUI polishing chnages (#40)
- 859ea18 add GitHub Copilot authentication provider (#37)
- f1a23a9 add duck duck go web_serarch (#41)
- 61e6833 add effort capability (#57)
- be1f64e add plan mode tool calling (#70)
- cc79d84 chore(tui): refactor command palette and fix long line rendering (#36)
- 6a79308 enhanse the git tools (#23)
- dc1a118 feat(folder trust): add trust command and first-launch prompt (#20)
- 7fef41c feat(tui): update path trust modal and tool card rendering (#86)
- 05c3084 fix summarize context overflow (#31)
- 042ba52 fix: improve copilot wizard and add documentation (#38)
- d7deeb0 fix: reclaim dispatch worktrees on every worker exit and at session teardown (#72)
- 17b0199 fix: sync AgentTool.Adapter on provider/model switches; add dock trailing blank (#69)
- fa233b1 format fixes (#48)
- 534a922 plan-mode polish: foreground parallelism, entry card, prompt nudges (#24)
- 3b99a1d polish tui output add better context format (#49)
- 0ba1db8 polish tui output add better context format (#53)
- c816925 polishing changes for /context (#47)
- 82455b4 refine prompts (#28)
- 7d6dfdb refine prompts (#80)
- 1d000aa refine the system messages format (#67)
- a872056 refine which provider has access to the image capability (#43)
- e3bbfd3 release: v0.3.0 Memory + ecosystem — persistent agent memory with semantic recall, installable skills, an MCP client, typed git/GitHub (#91)
- bc1f658 resolve empty return string bug (#77)
- 721aaf6 switch github tool to typed client, drop gh CLI (#19)
- 2bb51d2 tui format changes (#50)
- 98634de update models + enhance XAI provider (#82)
- ed56981 update readme docs (#66)
- aaefc24 update the readme file (#81)
v0.2.0
yottacode v0.2.0
Release Date: 2026-05-13
Control flow + safety triad — typed subagents, per-prompt checkpoints, plan & auto modes, and a custom-command starter kit.
✨ Highlights
- Typed subagents (
Agenttool +/subagentspicker) — delegate research, code search, and planning to typed subagents that run in their own context window. Three built-in types (general-purpose,Explore,Plan) plus custom.yottacode/agents/*.mddefinitions. Mode and approval flow forward into child runs (#5). - Per-prompt checkpoints (
/checkpoints/Esc Esc) — automatic file-snapshot rewind that mirrors Claude Code's/rewind. Every user message snapshots conversation + pre-edit file contents under~/.yottacode/checkpoints/<session>/, content-addressed and deduped with a 30-day TTL (#8). - Plan and auto modes —
/planenters a read-only design phase; approving with[Y]exits into auto mode for the implementation. Safety floor preserved:run_bash,git_commit,git_checkpoint, androllbackstill prompt. Mode state propagates into subagent runs (#4). todo_writetool + inline todo cards — model-driven plan tracking that mirrors Claude Code'sTodoWrite. The TUI renders each snapshot as a scrollback card; the list persists tosession.Todosand restores via/sessions <id>(#3).- Mid-turn interrupts —
Ctrl+Ccancels in-flight turns cleanly: streaming stops, the in-progress tool call receives a context cancellation, and conversation history is repaired so the next turn doesn't trip on a danglingtool_use. Read-only slash commands markedPreservesTurndon't cancel an active turn when run mid-flight (#7). - Custom slash commands + built-in starter kit — user/project commands loaded from
~/.yottacode/commands/and.yottacode/commands/. Four embedded defaults ship with the binary:/git:commit-message,/git:create-pr,/check:review,/check:verify. YAML frontmatter (description,argument-hint),$ARGUMENTS/$1..$9substitution, and@<path>file references supported (#10). - Experimental feature flag system —
--experimental,$YOTTACODE_EXPERIMENTAL, and[experimental]inconfig.tomlgate not-yet-stable features behind opt-in. First feature:background_subagents(off by default) (#5). - Permission file integrity checks —
permissions.jsonandpermissions.local.jsonare validated on load for shape, duplicate rules, and known rule-type prefixes. Malformed files surface a startup warning naming the offending entry rather than silently dropping rules (#11).
🐛 Bug Fixes & Improvements
- TUI
write_filecard no longer mis-renders when the target is a new file. Previously fell through to an edit-style diff with an emptybeforeblock; new files now render with a clear new-file indicator and full body preview (#9). - Built-in tool output rendering refactored in the TUI. Tool-result cards now compose from a per-tool renderer interface so future tools can ship with custom card shapes without touching the main TUI loop (#2).
Full Changelog: v0.1.0...v0.2.0