feat: user-provided reference/sketch images guide diagram generation#248
Merged
Conversation
Allow passing existing images (hand-drawn sketch, whiteboard photo, prior figure version) as guidance alongside the methodology text: - GenerationInput gains input_images: list[str] (paths) - Repeatable --image PATH on `paperbanana generate`; files are validated (existence + PIL-openable raster) before the pipeline starts, and rejected when combined with --continue/--continue-run - Planner attaches the images as additional image parts after the retrieved exemplar images, labeled as user-provided reference/sketch in the prompt (exemplar "reference image N" indexing is preserved) - Visualizer diagram prompt gets a one-line note that a user sketch guided the plan (carried via a sketch_guided flag so the note never leaks into the description the Critic reviews) - The Critic never sees the sketch: it keeps judging against the source text only - input_images survives the --optimize input rebuild and is recorded in run_input.json for reproducibility - MCP: generate_diagram gains optional input_images (validated paths) Fixes #223
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #223 —
--image PATH(repeatable) lets a hand-drawn sketch, whiteboard photo, or prior figure guide generation:GenerationInput.input_images: list[str]; validated pre-pipeline (exists + PIL-verifiable), rejected on continue runs--optimizeGenerationInput rebuild; recorded inrun_input.json; MCPgenerate_diagramgains the optionalinput_imagesparam with validation--optimizeis deliberately out of scopeSuite at 841 passing with the new coverage (planner attachment, critic blindness, CLI/MCP validation).