Latest Results
re add baml image inputs and outputs (#3481)
<!-- CURSOR_SUMMARY -->
> [!NOTE]
> **Medium Risk**
> Medium risk because it changes core parsing/lowering and LSP
diagnostics/completions behavior, which can affect editor features and
error reporting across many files. The changes are internal but touch
multiple compiler layers (parser → AST → HIR → visualization).
>
> **Overview**
> Improves declarative LLM ergonomics by threading `return_type` through
`PrimitiveClient.build_request*` calls and updating builtins
accordingly.
>
> Adds Jinja-aware validation for LLM prompts, `template_string`s, and
`@check`/`@assert` constraint expressions in `baml_lsp2_actions`, backed
by new `sys_jinja`/`sys_jinja_types` crates; includes span-accurate
diagnostics, cross-file template-string resolution, and updated golden
test expectations.
>
> Extends compiler2 lowering to collect `env.VAR` references (including
inside config blocks) into HIR, exposes them via a new
`all_env_var_names` API, and tightens LSP actions/annotations to skip
non-user/synthetic functions.
>
> Enhances editor/visualization UX: control-flow graphs gain an
`LlmFunction` node type plus optional `callee_name`/source span
metadata, visualization inlines non-empty branch-arm containers into
labeled edges, and completions now pull builtin primitive/media methods
from the stdlib (including static constructors and `uint8array`) while
improving parser recovery for malformed map/object literals and
header-comment LLM detection.
>
> <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit
6c5d2de18e080bd2529731cdac84065ce70bf5f9. Bugbot is set up for automated
code reviews on this repo. Configure
[here](https://www.cursor.com/dashboard/bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Image outputs: in-node previews, multi-image grids, and
image-generation provider support
* Persistent environment variables across sessions
* Source-code navigation from playground execution
* Toggle internal function visibility and origin badges in sidebar
* **Bug Fixes**
* Header comments containing "prompt" no longer misclassify LLM bodies
* Function execution events now include error details for clearer
diagnostics
* **Improvements**
* Richer graph visuals, layouts, and runtime previews
* Smarter media-aware prompt/request handling and enhanced completions
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: hellovai <vbv@boundaryml.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Latest Branches
+21%
-18%
avery/pattern-control-flow +67%
paulo/optional-function-parameters © 2026 CodSpeed Technology