Conversation
…t AI instruction files Migrate pipeline files from .plx to .mthds extension, restructure the inference deck into separate files (llm, img_gen, extract, custom), update backend configs, and consolidate AI coding instructions into CLAUDE.md by removing .blackboxrules, .cursor/rules, .windsurfrules.md, AGENTS.md, and .github/copilot-instructions.md. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace typing.List with builtin list[str] and remove unnecessary from __future__ import annotations to fix annotation resolution failure after pipelex update. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
| best-gemini = "nano-banana-pro" | ||
| best-blackforestlabs = "flux-2-pro" | ||
|
|
||
| default-general = "flux-2-pro" |
There was a problem hiding this comment.
Image generation alias references non-existent model
High Severity
The aliases best-blackforestlabs and default-general both point to "flux-2-pro", but no backend configuration file defines a model with that name. The fal.toml backend only defines flux-pro, flux-pro/v1.1, flux-pro/v1.1-ultra, flux-2, and fast-lightning-sdxl. Since default-general is used by the gen-image preset (which is the choice_default), all default image generation will fail to resolve a model. The intended model is likely flux-2.
| [extract.aliases] | ||
| default-premium = "azure-document-intelligence" | ||
| default-extract-document = "mistral-document-ai-2505" | ||
| default-extract-image = "mistral-document-ai-2505" |
There was a problem hiding this comment.
Extract alias references non-existent extraction model
High Severity
The aliases default-extract-document and default-extract-image both point to "mistral-document-ai-2505", but no backend defines this model. The Mistral backend only defines mistral-ocr-2503, mistral-ocr-2505, mistral-ocr-2512, and mistral-ocr. Since default-extract-document is the choice_default for extraction, all default document extraction will fail to resolve a model.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 94541fb8eb
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| @@ -28,7 +28,7 @@ description = "Find code files that implement or use elements mentioned in docs" | |||
| inputs = { doc_file = "DocumentationFile", repo_map = "RepositoryMap" } | |||
| output = "FilePath[]" | |||
| model = { model = "llm_for_large_codebase", temperature = 0.1 } | |||
There was a problem hiding this comment.
This pipe still uses model = "llm_for_large_codebase", but this commit deletes .pipelex/inference/deck/base_deck.toml (where that identifier was defined) and the new deck files (1_llm_deck.toml + x_custom_llm_deck.toml) do not reintroduce it, so model resolution for this step will fail at runtime (the same stale reference is also present in cocode/pipelines/text_utils.mthds).
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable autofix in the Cursor dashboard.
| output = "FilePath[]" | ||
| model = { model = "llm_for_large_codebase", temperature = 0.1 } | ||
| model_to_structure = "cheap_llm_for_structured" | ||
| model_to_structure = "@default-small-structured" |
There was a problem hiding this comment.
Deleted waterfall breaks existing pipeline model references
High Severity
Deleting base_deck.toml removes the llm_for_large_codebase waterfall, but two existing pipeline files (doc_proofread.mthds and text_utils.mthds) still reference it as a model name. The new deck files define no equivalent waterfall or alias, so these pipelines will fail to resolve the model at runtime.
1 participant
Note
Medium Risk
Touches core inference configuration (model decks and per-backend model capabilities), so misconfigurations could break model selection/routing or enable unsupported inputs (notably
pdf).Overview
Refactors Pipelex inference configuration by replacing the monolithic
base_deck.toml(plus old overrides/custom decks) with split deck files forllm,img_gen, andextract, and adds new custom-deck templates for advanced users.Updates many backend model definitions to include a new
thinking_modesetting, expands several LLMs to acceptpdfinput, adds/updates model entries (e.g., newer Claude/GPT/Gemini/Mistral variants), and introduces “positive-only” prompt rules for multiple image-generation backends; also renames the internal PyPDFium extractor topypdfium2-extract-pdf.Written by Cursor Bugbot for commit 7ed82a7. This will update automatically on new commits. Configure here.