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.

Fix docs and URLs and add Changelog

Louis Choquel861d409e21f0
diff --git a/CHANGELOG.md b/CHANGELOG.md
index e57c5b0e3..18f4c5b97 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,8 +2,26 @@
## Unreleased
-- Updated the api documentation.
-- Updated the routing profiles.
+### Added
+ - Model Waterfalls: Define prioritized model lists in `base_deck.toml` (e.g., `smart_llm = ["gpt-4o", "claude-4.5-sonnet", "grok-3"]`). Pipelex automatically falls back to the next model if the preferred one is unavailable.
+ - Advanced Routing Profiles: New capabilities in `routing_profiles.toml`: `fallback_order` (Global fallback sequence specifying which backends to try if a model isn't found) and `optional_routes` (Routes that activate only when their target backend is enabled)
+ - New Models: Anthropic `claude-4.5-haiku` (Pipelex Inference, Anthropic, and Bedrock backends) and Azure OpenAI `o3`
+ - Comprehensive Documentation Relaunch: Complete restructure under `/home/` with new "Get Started" guides for `pipelex build` and manual workflows, plus in-depth sections on Domains, Bundles, Concepts, and Pipe lifecycle.
+ - Enhanced CLI: `pipelex init` now interactively guides backend selection and automatically configures routing profiles, including primary backend and fallback order. Added `pipelex init routing` focus.
+ - Enhanced CLI: Improved error reporting across all commands (`build`, `validate`, `run`, `show`) with clear, actionable feedback for configuration errors, missing models, and invalid presets.
+ - Enhanced CLI: `pipelex doctor` now validates model deck configuration.
+ - New Routing Profiles: Full suite of `all_*` profiles (e.g., `all_openai`, `all_anthropic`, `all_google`) to route all requests to a single provider.
+
+### Changed
+ - BREAKING: for inline concept structures, the fields are now optional by default: the `required` property defaults to `false`. Explicitly set `required = true` to make fields mandatory, which we discourage as it increases risks of hallucinations.
+ - LLM Presets Overhaul: Rationalized and renamed default presets in `base_deck.toml`. Single-model aliases replaced with waterfall aliases. Key renames: `llm_for_complex_reasoning` → `llm_to_engineer`, `llm_to_answer_hard_questions` → `llm_to_answer_questions`, `llm_to_write_questions` → `llm_for_writing_cheap`. Removed redundant older presets.
+ - Stricter Configuration Validation: Pipelex validates model deck on startup and raises errors if presets reference unavailable models.
+
+### Fixed
+ - Local OpenAI-Compatible Endpoints: OpenAI plugin now handles empty API keys, enabling seamless integration with local servers like Ollama.
+
+### Removed
+ - Old Documentation Structure: Previous `/pages/` directory documentation removed in favor of new structure.
## [v0.14.3] - 2025-10-29
diff --git a/docs/home/5-setup/configure-ai-providers.md b/docs/home/5-setup/configure-ai-providers.md
index bc016b940..d390a8290 100644
--- a/docs/home/5-setup/configure-ai-providers.md
+++ b/docs/home/5-setup/configure-ai-providers.md
@@ -11,7 +11,7 @@ Get **free credits** for testing and development with a single API key that work
**Benefits:**
- No credit card required
-- Access to OpenAI, Anthropic, Google, Mistral, and more
+- Access to OpenAI, Anthropic Claude, Google Gemini, and xAI Grok
- Perfect for development and testing
- Single API key for all models
diff --git a/pipelex/cogt/models/model_deck.py b/pipelex/cogt/models/model_deck.py
index 358820cac..840db84f0 100644
--- a/pipelex/cogt/models/model_deck.py
+++ b/pipelex/cogt/models/model_deck.py
@@ -382,7 +382,7 @@ class ModelDeck(ConfigModel):
"'.pipelex/inference/backends.toml' and that you have the necessary credentials."
"To find what backend is required for this model, look at the routing profile in .pipelex/inference/routing_profiles.toml. "
"Learn more about the inference backend system in the Pipelex documentation: "
- "https://docs.pipelex.com/pages/configuration/config-technical/inference-backend-config/"
+ f"{URLs.backend_provider_docs}"
)
raise ModelNotFoundError(message=msg, model_handle=model_handle)
diff --git a/pipelex/urls.py b/pipelex/urls.py
index 67eef86f5..09674bdc9 100644
--- a/pipelex/urls.py
+++ b/pipelex/urls.py
@@ -4,5 +4,5 @@ class URLs:
documentation = "https://docs.pipelex.com/"
changelog = "https://docs.pipelex.com/changelog/"
discord = "https://go.pipelex.com/discord"
- pipe_func_docs = "https://docs.pipelex.com/pages/build-reliable-ai-workflows-with-pipelex/pipe-operators/PipeFunc"
- backend_provider_docs = "https://docs.pipelex.com/pages/configuration/config-technical/inference-backend-config/"
+ pipe_func_docs = "https://docs.pipelex.com/home/6-build-reliable-ai-workflows/pipes/pipe-operators/PipeFunc/"
+ backend_provider_docs = "https://docs.pipelex.com/home/5-setup/configure-ai-providers/"