npm @earendil-works/pi-coding-agent 1.6.0
v1.6.0

one hour ago

New Features

  • GPT-6.2 Astra and GPT-6.2 Sol: both models are available on the openai, azure, and openai-codex providers, and GPT-6.2 Sol is now the default model for all three. See Providers.

Breaking Changes

  • Extension components: render() now returns readonly string[], and Container returns the same cached array while its children are unchanged. Extensions that modify render output in place, including super.render() in a Container subclass or an embedded component's output, must copy it first (#9549)

Added

  • Added GPT-6.2 Astra and GPT-6.2 Sol to the openai, azure, and openai-codex providers. Both reject reasoning effort none and minimal, so thinking level off is unavailable. Azure serves them with a 1.05M context window; Codex serves GPT-6.2 Sol with 500K and GPT-6.2 Astra with 272K
  • Added qwen4-max-preview to the Qwen Token Plan Individual providers (#12259)
  • Added deepseek-v4.1-flash and deepseek-v4.1-flash-1017 to the Qwen Token Plan Individual providers, and per-plan key variables QWEN_TOKEN_PLAN_TEAM_API_KEY, QWEN_TOKEN_PLAN_TEAM_CN_API_KEY, QWEN_TOKEN_PLAN_INDIVIDUAL_API_KEY, and QWEN_TOKEN_PLAN_INDIVIDUAL_CN_API_KEY; the shared QWEN_TOKEN_PLAN_API_KEY and QWEN_TOKEN_PLAN_CN_API_KEY still work (#11773)
  • Added a transcript notice when OpenAI drops reasoning items from a request, such as OpenAI dropped 2 reasoning items: prefix_mismatch (details in session) after compaction on GPT-6.2. Like the Anthropic thinking notice, it is shown when showCacheMissNotices is on and only when more items are dropped than in the previous response. Both notices now name the reasons

Changed

  • Qwen Token Plan providers now use the current gateways (token-plan.maas.qwencloudapi.com international, token-plan.maas.qianwenaiapi.com China), and qwen-token-plan and qwen-token-plan-cn are named "Qwen Token Plan Team" and "Qwen Token Plan Team CN". Provider IDs are unchanged, and the old aliyuncs.com hosts still work (#11773)
  • Updated undici to 8.12.4 for security advisories GHSA-mqg3-3v4r-cm32, GHSA-7xg3-fphp-28hg, GHSA-5rm8-gv32-r954, GHSA-9v6h-8rgp-m69f, GHSA-3g9m-rv3r-w5w2, and GHSA-r4pc-4jcp-7pqc (#12200)
  • The default model for the openai, azure, and openai-codex providers is now GPT-6.2 Sol (gpt-6.2-sol) instead of GPT-6.1 Sol, at the same price
  • The find tool's path patterns are now relative to the search directory, like grep's glob: src/*.ts finds src/a.ts but no longer pkg/src/a.ts; use **/src/*.ts to match at any depth. Patterns without / still match file names at any depth (#12066)
  • The llama.cpp extension now connects to http://127.0.0.1:9931 by default, the current llama-server default port, instead of port 8080 (#12001 by @julien-c)
  • Anthropic message threads are now on by default for any provider whose baseUrl is on api.anthropic.com, not only the anthropic provider (#11987)
  • The resume command printed on exit is now on its own line after To resume this session:, so it can be selected with a triple-click (#12237)

Fixed

  • Fixed read storing a note that the current model does not support images, which was replayed to the model as stale text after switching to a vision model. The image is still replaced with a placeholder per request for models without image input, and the read result now shows the warning in the TUI only, also when collapsed (#12131)
  • Fixed MCP tools loaded by tool_search being dropped on resume and /reload when an extension deactivates only its own tool at session_start, as pi-web-access and pi-subagents do in toolActivation: "auto" mode (#11996)
  • Fixed the standalone binary failing to resolve extension npm dependencies whose entry point is declared in package.json main or exports (for example shiki, yaml, graceful-fs), including imports made from inside other dependencies (#12012)
  • Fixed the find tool returning no results for patterns starting with ./, such as ./src/*.ts (#12020)
  • Fixed pi exiting with an uncaught AbortError when a read tool call is cancelled while the file is being opened (#11956)
  • Fixed pi exiting with an uncaught AbortError when the session picker cancels loading after skipping a .jsonl file that is not a session, and fixed such files staying open (#11956)
  • Fixed /share crashing pi when the GitHub CLI fails to start
  • Fixed pi update on Windows managed installs failing with "Could not verify managed Pi" when an older system Node.js comes before the installer's standalone Node.js on PATH (#12048)
  • Fixed /export HTML showing raw terminal hyperlink escape codes around links in custom tool output; links are now clickable (#12003)
  • Fixed the bash tool returning no output, with a footer like [Showing lines 3-2 of 2], when the output ends in one line longer than about 100KB, such as a grep match in a minified file; it now shows the end of that line (#12062)
  • Fixed provider errors reading "Connection closed." not triggering automatic retries (#12064)
  • Fixed /resume sometimes selecting an unrelated session while the list reloads after deleting or renaming a session (#12106)
  • Fixed Groq Qwen models failing with 400 Unexpected message role; OpenAI-compatible providers other than OpenAI now receive system messages (the system prompt and mid-conversation system messages) with the system role (#10741)
  • Fixed enabledModels/--models warnings such as No models match pattern "github-copilot/nope" being written over the interactive TUI on every /new, /resume and /fork, which left stale lines in fullscreen mode. The warnings are now shown once at startup, in the chat in interactive mode and on stderr in other modes; RPC mode no longer repeats them on stderr for each new_session, switch_session or fork (#11264)
  • Fixed the find tool matching path patterns through folders above the search path: with the project in ~/src/app, src/**/*.ts returned every .ts file in the project (#12066)
  • Fixed pi crashing with RangeError: Maximum call stack size exceeded when rendering a very large edit diff, for example after editing a file with a multi-megabyte line, and again on every resume of that session (#8036)
  • Fixed pressing Esc during a batch of tool calls leaving the calls that had not started yet without a result in the session; they now get an "Operation aborted" result (#9986)
  • Fixed an unreadable CHANGELOG.md (pi's own or an extension package's) printing Warning: Could not parse changelog over the interactive TUI; it now shows no changelog notes (#11264)
  • Fixed problems found when replacing the session with /new, /resume, /fork or RPC new_session, switch_session and fork never being shown, for example a project extension that fails to load after /resume into another project, or an enabledModels pattern that matches nothing after a settings change. Interactive mode now shows them in the chat and print, JSON and RPC modes write them to stderr; a problem already reported is not repeated. SDK hosts can call the new AgentSessionRuntime.takeUnreportedDiagnostics() at startup and after each replacement to get each diagnostic once
  • Fixed RPC new_session, switch_session, fork and clone running extension session_start and resources_discover handlers twice for the new session. RPC clients now also receive events emitted by session_start handlers, such as entry_appended for entries they append (#10470)
  • Fixed an exception thrown by an extension's pi.events handler being printed with console.error over the interactive TUI. It is now reported like other extension errors: in the chat in interactive mode, as extension_error in RPC mode and as Extension error (...) in print mode. Errors from a handler that fails after its session was replaced are dropped (#11264)
  • Fixed AgentSession.reload() not emitting session_start to extensions when the SDK host bound them with bindExtensions({}), after it had already emitted session_shutdown. Extensions that start work in session_start, such as the MCP extension connecting its servers, stayed stopped after a reload
  • Fixed --mode json dropping events emitted by extension session_start handlers, such as entry_appended for entries they append, at startup and after an extension replaces the session
  • Fixed Could not restore model <provider>/<model>. Using <fallback> being shown only at interactive startup. Resuming or forking into a session whose model no longer exists or has no login switched to another model without notice, and print, JSON and RPC modes never showed it. It is now reported in every mode, at startup and after /resume, /fork and RPC switch_session or fork
  • Fixed git and npm output, such as Cloning into ... or added 12 packages, being drawn over the interactive TUI when pi installed or refreshed packages during /reload, /new, /resume and /fork, including temporary -e git: and -e npm: packages. In interactive mode their output is now captured, and a failed install reports the end of it. Until the TUI starts, each install is announced as Installing <source>... on stderr instead of showing git and npm output. git and ssh can no longer prompt in interactive mode: a package that needs a typed password or passphrase fails with a hint to set up a credential helper or ssh agent, or to run pi install <source> once from the shell. pi install and pi update keep live output (#11264)
  • Fixed fullscreen scrolling in long sessions saturating a CPU core by re-rendering the whole transcript on every wheel event (#9549)
  • Fixed search in /login, /logout, /scoped-models and /resume keeping the highlight on the row scrolled to before typing, so Enter picked whichever entry landed there instead of the best match; a query now highlights the top match, as /model does (#12245, #12246 by @acmerfight)
  • Fixed exiting pi, pi config and the --resume and other startup pickers dropping terminal output still queued in stdout, such as the resets that turn off mouse tracking and Kitty keyboard reporting; the shell could then print mouse and key reports as text. pi now waits up to one second for queued output before exiting or continuing (#12133)
  • Fixed moving the cursor in a search field with Left, Right, Home or End moving the highlight back to the first entry in /login, /logout, /model, /scoped-models, /resume, /settings, the per-model settings model picker and pi config (#12245)
  • Fixed llama.cpp decision classifiers ignoring the image input that llama.cpp reports, such as Clef with its mmproj; models.classify() now sends their images to /v1/systemone (#12282)
  • Fixed kill -TSTP from outside pi stopping it without restoring the terminal, which left mouse tracking and the Kitty keyboard protocol on at the shell prompt and turned mouse movement into garbage input. pi now suspends as with Ctrl+Z and redraws on fg (#12283 by @haoqixu)
  • Fixed changing the thinking level on the openai-codex provider invalidating the prompt cache on GPT-6 models (#12269)
  • Fixed unscoped syntax-highlighted text inheriting the terminal foreground and becoming unreadable on contrasting tool cards (#11092)
  • Fixed /copy saying "No agent messages to copy yet." after Esc during a tool call or after a provider error; it now copies the previous answer. RPC get_last_assistant_text skips the same messages (#12119, #12121 by @acmerfight)
  • Fixed pi.exec reporting exit code 0 for commands ended by a signal, such as a timeout, an abort or a crash; it now reports 128 + the signal number, like a shell (143 for SIGTERM, 137 for SIGKILL) (#12088, #12089 by @acmerfight)
  • Fixed GitHub Copilot Gemini and Kimi models (gemini-4-argon, gemini-4-carbon, gemini-3.8-flash, kimi-k3) ignoring the thinking level, and Copilot usage always reporting 0 reasoning tokens (#12110 by @DenielC1)
  • Fixed provider errors that report only a Node socket error code, such as read ECONNRESET, write EPIPE, or connect ETIMEDOUT, not triggering automatic retries (#12128)
  • Fixed OpenAI rejecting requests with a 400 error when a tool's schema contains a node without a type (such as items: {} or Type.Unknown()); such tools are now sent in non-strict mode (#12049)
  • Fixed Mistral Large 4 tool calls stalling for minutes; Mistral models no longer receive strict tools unless compat.supportsStrictMode is set (#12148)
  • Fixed non-strict Anthropic tool schemas not telling the model that extra arguments are forbidden (#12067)
  • Fixed GitHub Copilot models on the Responses API (GPT, Grok, MAI-Code) filling omitted optional tool arguments with empty strings or invented values (#12286)
  • Fixed the OpenAI Codex WebSocket 60-minute connection limit failing the turn when it hit mid-stream, for example during long thinking; the request is now retried on a fresh connection (#12124)
  • Fixed the openai-codex provider getting almost no prompt cache hits for the system prompt on GPT-6.2 models (#12273)
  • Fixed the openai-codex provider ignoring a model's maxTokens
  • Fixed fullscreen Kitty images disappearing after a full redraw such as a window resize (#12010)
  • Fixed text after a multi-row inline image in Markdown drawing over the image (#12017)
  • Fixed Kitty and iTerm2 images covering overlays drawn over them (#11584)
  • Fixed Enter on a Tab path menu for an absolute path (look at /tmp/fi) or a slash command argument menu (/open /tm) submitting the prompt instead of accepting the completion (#12309, #12312 by @acmerfight)

Removed

  • Removed the retired deepseek-v3.2 and deepseek-v4-flash from the Qwen Token Plan Team providers; use deepseek-v4-flash-0731 or deepseek-v4.1-flash (#11773)

Don't miss a new pi-coding-agent release

NewReleases is sending notifications on new releases.