Skip to content

Build AI methods with your coding agent, then run them anywhere: as an MCP for chatbots, as a webapp for people, or via API for your software.

Changelog and test count

Louis Choquelbbcf532d0134
diff --git a/.badges/tests.json b/.badges/tests.json
index ea3a06bef..25e249817 100644
--- a/.badges/tests.json
+++ b/.badges/tests.json
@@ -1,7 +1,7 @@
{
"schemaVersion": 1,
"label": "tests",
- "message": "1526",
+ "message": "1572",
"color": "blue",
"cacheSeconds": 300
}
\ No newline at end of file
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2bde10241..d9c7298a1 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,6 @@
# Changelog
-## Unreleased
+
+## [v0.15.5] - 2025-11-18
### Added
- **Improved configuration repair**: New `--fix` option for `pipelex doctor` command that interactively detects and repairs outdated or invalid backend configuration files using latest templates from the Pipelex kit.
@@ -7,12 +8,15 @@
- **Enhanced test infrastructure**: Integration tests now automatically parameterized to run against all supported backend routing profiles. Tests are intelligently skipped at collection time if a model is not supported by the active backend profile, with a summary of skipped tests provided at session end.
- **New routing profiles**: Added `all_groq` and `all_pipelex_inference` routing profiles.
- **Vision support flag**: Added `is_vision_supported` property to `LLMWorkerAbstract` class for explicit checks of model vision capabilities.
+- **New type of `StuffContent`**: `JSONContent` to support an arbitrary JSON object as input or output of a pipe.
+ - **Azure image generation**: Support for image generation models via Azure OpenAI backend using `gpt-image-1`.
### Changed
- **Unified structured output**: Complete overhaul of structured generation settings. Replaced global configuration setting with new `structure_method` parameter in backend `.toml` files (configurable at backend level in `[defaults]` or per individual model). Expanded `StructureMethod` enum to include dozens of modes supported by `instructor`, enabling fine-grained control over provider-specific features like OpenAI Structured Outputs or Anthropic Tools, and various JSON-based modes.
- **Groq integration**: Updated to use standard `openai` SDK, simplifying integration.
- **Default configurations**: All official backend providers now enabled by default after `pipelex init`. Default prompting style changed from `ticks` to `xml`.
- **Code & test organization**: Unit test suite reorganized from `tests/unit/core` and other directories into unified `tests/unit/pipelex/` structure. Integration test fixtures modularized from `conftest.py` into separate files within `tests/integration/pipelex/fixtures/`.
+ - **Console output settings**: Added `console_print_target` and `console_log_target` settings in `pipelex.toml` for redirecting output to `stdout` or `stderr`, with CLI and logging refactored to use centralized console instance. This makes it easier to support MCP communication based on stdio.
### Fixed
- Improved Pydantic validation error messages when loading backend configurations to clearly indicate the specific file and model containing the error.
@@ -22,12 +26,6 @@
- **Groq plugin**: Dedicated `pipelex/plugins/groq` plugin removed (now uses standard `openai` SDK).
- **Global instructor config**: Global `is_openai_structured_output_enabled` setting, replaced by per-model `structure_method` approach.
-## Unrelease
-
-### Added
-
-- JSONContent to have as inputs and outputs of pipe unknown JSON format
-
## [v0.15.4] - 2025-11-12
### Added