Skip to content

Version v0.9.9.5

Latest
Latest

Choose a tag to compare

View all tags
@karthink karthink released this 04 May 05:55
· 2 commits to master since this release
v0.9.9.5
f5ad4eb

Version 0.9.9.5 adds support for new OpenAI, Anthropic, Gemini, Copilot and xAI models, the OpenAI Responses API, token usage display and better UI indicators, tool call hooks, inspection and interactive editing, a default preset and more.

Breaking changes

  • gptel’s default ChatGPT backend has been removed. gptel-backend and gptel-model now default to nil, and there are no registered backends out of the box. However gptel remains usable without configuration: if gptel-send is called without a backend set, the ChatGPT backend is created on the fly and used.

  • gptel-track-media affects gptel’s link handling in all Org and Markdown buffers, not just chat buffers that have gptel-mode turned on. When calling gptel-send with gptel-track-media turned on, and the buffer is in Org or Markdown mode, links to supported file types will be followed by gptel and included with the request. Previously this behavior applied only in dedicated chat buffers.

    (This is actually how gptel has worked since v0.9.9.3, but this change in behavior was undocumented.)

  • gptel-include-reasoning now defaults to ignore, meaning that reasoning text in LLM responses will be included in buffers but ignored by gptel-send on subsequent conversation turns. The reason for this change is that including the reasoning text as the LLM’s response on new conversation turns is not recommeded by LLM APIs. Reasoning text can also fill up the context window.

  • gptel-make-tool now sets the tool’s :include slot by default. This means that unless :include nil is explicitly specified, gptel-tools will default to including their results in the buffer when using gptel-send. This is recommended for coherent multi-turn conversations involving tool use, as the LLM uses tool results from past turns for context.

    (Tool result inclusion can be controlled globally (or buffer-locally) for all tools via gptel-include-tool-results, whose default value has not been altered.)

  • The models gpt-41-copilot, gpt-5 and claude-opus-41 have been removed from the default list of GitHub Copilot models. These models are no longer available in the GitHub Copilot API.

  • The models gpt-3.5-turbo and gpt-3.5-turbo-16k have been removed from the default list of OpenAI models. These models are either deprecated or no longer available.

  • Breaking change to the gptel-backend API: When the header or url fields of a backend are specified as functions, they now accept one argument: the request context plist (info). This is only relevant if you have defined custom header or url functions for your gptel backends.

    (This change is required for backends whose request header or URL should be modified depending on the state of the request.)

New models and backends

  • Deepseek backend: Add support for deepseek-v4-flash, deepseek-v4-pro.

  • Anthropic backend: Add support for claude-opus-4.7.

  • xAI backend: Add support for grok-4-1-fast-reasoning, grok-4-1-fast-non-reasoning, grok-4-fast-reasoning, and grok-4-fast-non-reasoning.

  • GitHub Copilot backend: Add support for claude-opus-4.7, gpt-5.1-codex-, =gpt-5.1-codex-mini, =claude-sonnet-4.6, gemini-3.1-pro-preview, gpt-5.3-codex, gpt-5.4, gpt-5.4-mini, and gpt-5.5.

  • Gemini backend: Add support for gemini-3.1-flash-lite-preview; add deprecation notice for gemini-3-pro-preview.

  • OpenAI backend: Add support for gpt-5.5, gpt-5.5-pro, gpt-5.3-chat-latest, gpt-5.4, gpt-5.4-pro, gpt-5.4-mini, gpt-5.4-nano. gpt-5.2, gpt-5-mini, gpt-5-nano and o3-pro.

New features and UI changes

  • gptel now displays token usage in the header-line when using gptel-mode. This includes the tokens sent, cached tokens sent (if any) and tokens received. The displayed totals are per-request or per session/buffer, and you can switch between the two by clicking the display. Hovering the mouse over the display will show both in a tooltip.

    Displaying costs in currency instead of tokens is not yet supported because many backends do not provide all the required information (tiered pricing schemes, cache creation details, etc).

  • New command gptel-preset to choose a gptel preset to apply, or save gptel’s current settings as a preset.

    gptel-menu now uses this command instead of a transient menu for applying presets.

  • gptel now ships with a pre-defined preset named gptel-default. This preset “resets” most gptel settings to the default values that gptel ships with. This means no system prompt or tools, no additional context sources, no request temperature or max token limits and more. It does not change the active gptel-model or gptel-backend. This has at least two uses:

    1. By applying the preset interactively, you can use it to quickly reset most of gptel’s configuration back to the default (excepting the model and backend, which are nil by default).

    2. When using gptel-request for scripting purposes, you can fully isolate the call from the effects of gptel’s environment with

           (gptel-with-preset 'gptel-default
       (gptel-request prompt ...))

      Presets can also inherit from gptel-default to build a bespoke gptel configuration. (See gptel-make-preset for details.)

  • The UI indicators in chat buffers that report status changes (such as Waiting, Ready etc) now show the names of tools being called. The mode-line indicator display (see gptel-use-header-line) has been improved.

  • gptel-rewrite now displays status updates above the region being rewritten, indicating the request state (Waiting, Typing etc) and any tool calls in progress. This provides visual feedback for rewrites in progress, which was previously completely absent until the response was complete.

  • When using setopt or the customize interface, gptel-backend can now be specified as a list instead of an opaque object. See its documentation for details.

  • When using gptel-send, tool calls that require confirmation can now be examined in full in a dedicated inspection buffer, where they are displayed as Elisp forms.

    The tool name and tool call arguments can also be modified in-place now. These modifications must be in-place; deleting tool calls or adding new ones to the inspection buffer is not supported.

  • New hooks gptel-pre-tool-call-functions and gptel-post-tool-call-functions run before and after each tool call, respectively. These hooks receive details of the (planned or finished) tool call and provide fine-grained control over them. These hooks work with gptel-send, including when invoked from gptel’s Transient menu or from Elisp.

    gptel-pre-tool-call-functions can be used to modify tool call arguments, short-circuit the call and provide the results, block the tool call but continue the request with a message for the LLM, or stop the request entirely.

    gptel-post-tool-call-functions can be used to modify tool call results, block the tool call but continue the request, or stop the request entirely.

  • New variable gptel-bedrock-aws-cli-command to set the path to the AWS CLI command for the Bedrock backend. Defaults to “ews”.

  • gptel now supports the OpenAI Responses API. Practically speaking, you should notice no change in behavior, except that the newest OpenAI models should now be supported and available in OpenAI backends.

    Thinking/Reasoning in responses is also supported for OpenAI models now, but this is not yet configurable via gptel-include-reasoning.

Notable bug fixes

  • Authentication keys are now supplied to Curl (when gptel-use-curl is set) via STDIN instead of the command line. This is more secure as the command line is usually readable system-wide.

  • gptel-backend can now be set from customize buffers. These are produced by, for example, M-x customize-group ⮐ gptel. Previously gptel-backend was displayed in a read-only way, and could even break the display of the customize buffer depending on its value.

  • Breaking change to the gptel-request API: Tool call arguments are passed to gptel-request callbacks as a plist, not a list. The plist keys are the function argument names as specified in the tool definition. This does not affect gptel-send or (to my best knowledge) any of the packages using gptel.

    Example: The previous behavior was

    (funcall callback `(tool-call ,web-search ("emacs" 10) ,tool-cb))

    where web-search is a gptel-tool, and callback and tool-cb are the gptel-request and tool callback respectively. The new behavior is:

    (funcall callback
         `(tool-call ,web-search (:query "emacs" :count 10) ,tool,cb))

    where :query and :count correspond to arguments query and count in the definition of web-search.

    Note that this is a bug fix. It’s how the API has been documented and was supposed to work in the first place.

What's Changed

  • xAI: grok-4{,-1}-fast-{,-non}-reasoning by @endgame in #1269
  • Fix header-line right-alignment for variable-pitch fonts by @kovan in #1268
  • gptel-openai-extras: Correct context window limits on xAI models by @endgame in #1270
  • gptel-gh: Update models list by @gessen in #1271
  • gptel-gemini: add support for gemini-3.1-flash-lite-preview, deprecate gemini-3-pro-preview by @surenkov in #1278
  • Add tool call editing and tool call hooks by @karthink in #1261
  • feat: add customizable AWS CLI command path by @liangfu in #1251
  • Fix to allow no-argument tools to work with Bedrock by @jhoodsmith in #1265
  • gptel-openai: Add gpt-5.3-chat-latest and gpt-5.4 models by @berenddeboer in #1286
  • #1289, Groq example config creates ‘invalid_request_error’ by @andreas-roehler in #1290
  • #1294, fix: Deprecated model ‘llama3-70b-8192’ removed by @andreas-roehler in #1295
  • Fix search-failed error when curl exits without completing transfer by @benthamite in #1293
  • gptel-openai: Correct gpt-5.4 pricing and context window by @pabl0 in #1304
  • gptel: Ensure safety of file local variables by @pabl0 in #1306
  • gptel-org: Fix bug in link validation by @pabl0 in #1319
  • gptel-gemini: Fix temperature parameter mapping (#1312) by @Jeremias-A-Queiroz in #1313
  • Improve security of authentication with Curl by @pabl0 in #1308
  • Give crowdsourced prompts some love by @pabl0 in #1307
  • fix(transient): Resolve history symbol for read-number (#1314) by @Jeremias-A-Queiroz in #1315
  • Add gpt-5.4-mini and gpt-5.4-nano by @djr7C4 in #1316
  • Add Responses API with Copilot support by @karthink in #1292
  • Add gpt-5.4-pro and o3-pro models by @djr7C4 in #1323
  • update Github copilot endpoints docs by @CsBigDataHub in #1285
  • Treat gpt-5.4-pro and o3-pro as reasoning models in gptel--request-data by @djr7C4 in #1327
  • Fix nil token fields in message_delta crashing streaming tool calls by @parsnips in #1331
  • gptel: allow anonymous plist parents in gptel--apply-preset by @kmontag in #1332
  • Fix streaming responses by @Azkae in #1339
  • Ensure that gptel--ediff-restore runs after ediff-cleanup-mess by @djr7C4 in #1354
  • Fix the name of the gptel-preset transient in gptel-rewrite (fix #1350) by @djr7C4 in #1352
  • ensure newline between text and tool by @Azkae in #1353
  • Fix crash in gptel-mcp--activate-tools when requested servers return no tools by @benthamite in #1321
  • gptel: fix gptel--preset-syms dropping parent-contributed symbols by @kmontag in #1355
  • gptel-gh: Add claude-opus-4.7 to model list by @Jonghyun-Yun in #1361
  • Add Opus 4.7 as an Anthropic model. by @djr7C4 in #1362
  • gptel-request: Add -N flag to curl on all platforms to fix regression by @aagit in #1363
  • Add gpt-5.5 and gpt-5.5-pro by @djr7C4 in #1376
  • gptel-transient: Update the transient menu after applying prefixes by @djr7C4 in #1374
  • gptel-preset: allow system directives to be nil by @Stebalien in #1369
  • Add new DeepSeek backend configurations by @beacoder in #1377
  • Update NEWS with new deepseek backend model support. by @beacoder in #1379
  • gptel-openai-extras: update costs for deepseek-chat and deepseek-reasoner by @djr7C4 in #1381
  • gh: add gpt-5.5 to Copilot models by @Jonghyun-Yun in #1385
  • fix: Ensure utf-8 in tool call args by @FrauH0lle in #1392

New Contributors

Full Changelog: v0.9.9.4...v0.9.9.5

Contributors

kovan, endgame, and 20 other contributors
Assets 2
Loading