1 Version 1.27.2
════════════════

  • Fixed `make' recipes when a sandboxed caller inherits a stale
    `TMPDIR'.  Ellama now keeps valid temporary directories and falls
    back to `/tmp' only when the inherited directory no longer exists,
    which prevents Org from failing while loading during `srt make
    check-elisp'.


2 Version 1.27.1
════════════════

  • Added copyright notice to `extract-commands.el' script.


3 Version 1.27.0
════════════════

  • Added `ellama-undo-on-error'. Users can now choose whether failed
    LLM requests roll back partial buffer changes or leave the generated
    text in place.
  • Made plan-and-act and subagent loops more resilient after LLM
    failures. The loops continue after the first error, compact the
    session after repeated errors, resume after successful compaction,
    and stop cleanly when compaction cannot recover the session.
  • Made subagent tool errors part of the conversation instead of
    aborting the provider tool-use path. Synchronous and asynchronous
    tool failures are now returned as tool results so the worker can
    inspect the error and keep going.
  • Added preflight session compaction before oversized requests are
    sent. This covers regular `ellama-stream' calls and plan-and-act
    controller turns, and it resumes the deferred request in the target
    session buffer after compaction.
  • Changed edit tools to use asynchronous hook handling. Hooks no
    longer block the Emacs UI, while agents still receive the tool
    result only after the hooks have finished.
  • Tightened project checks. Byte and native compiler warnings now fail
      `make check-elisp', README custom-variable coverage is checked,
      and documented public commands are checked against real
      interactive commands.
  • Added configurable trusted read-file handling for
    DLP. Prompt-injection scans can be skipped for safe file-reading
    outputs without disabling DLP for other tools.
  • Improved agent buffer updates. Agent prompts, controller
    continuations, plan reports, and visible tool/status output now
    trigger the same auto-scrolling behavior as normal chat insertions,
    and finished plan-and-act loops append the next user heading.
  • Extended interactive model editing beyond Ollama and
    OpenAI-compatible providers. The model transient can now read and
    update providers with a generic `chat-model' slot, and URL editing
    is available for providers with a `url' slot.
  • Refreshed the documentation for the current agent, DLP, compaction,
    provider, skill, blueprint, and tooling behavior. The
    chat-with-image transient key was also moved from `i' to `I' to
    avoid a key conflict.


4 Version 1.26.0
════════════════

  • Auto-fix unexpected closing delimiters in file edits. When syntax
    validation fails due to unexpected closing delimiters (like extra
    `)` or `]`), the system automatically removes them and accepts the
    edit if the result is valid. This reduces friction for LLM-generated
    edits that occasionally include spurious closers. Changes include
    new functions `ellama-eval–remove-unexpected-closers-in-buffer` and
    `ellama-tools–remove-unexpected-closers-in-buffer` to scan buffers
    and remove closers without matching openers,
    `ellama-eval–try-fix-unexpected-closers` and
    `ellama-tools–try-fix-unexpected-closers` to attempt auto-fixing in
    file context with safe buffer restoration, and integration of
    auto-fixing into monitored-edit-file-tool and all balanced edit
    tools.
  • Add commit patch fallback for generate-commit-message. New functions
    `ellama–in-commit-buffer` and `ellama–commit-patch` detect git
    commit buffers and retrieve commit patches. Modified
    `ellama-generate-commit-message` to use commit patch as a fallback
    when no staged or unstaged diff is available, enabling proper commit
    message generation in git-commit-mode reword or amend scenarios.
  • Simplify test target in Makefile. Remove the inline test command
    definition and make `test` an alias for `test-detailed`,
    consolidating the test execution into a single target.


5 Version 1.25.0
════════════════

  • Add loop detection to `ellama-eval' harness that identifies
    repetitive tool
  …  …
