fix(ingester): handle malformed input, support source mapping #72
Conversation
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
- Added parsing of special Sources blocks delimited by `---` … `Sources:` … `---`.
- Computes active source ranges; assigns `chunk.meta.sourceLink` to the first URL listed in the most recent Sources block for all subsequent chunks until the next block.
- Extended `ChunkMeta` with optional `sourceLink` and tokenizer `Tokens` with `sourceRanges`.
- Wired sourceRanges into metadata attachment without altering chunking behavior.
- Tests
- New test: `ingesters/src/utils/__tests__/RecursiveMarkdownSplitter.sources.test.ts` validates activeSource mapping across multiple Sources blocks.
- Robust fenced code detection
- Accept up to 3 leading spaces before fences.
- Track fence char and exact length; closing fence must match char and have length ≥ open, fence-only (no info string).
- Fallback-close a malformed open block when a new opening fence appears while still in a block (marks previous as breakable).
- Mark unclosed-at-EOF blocks as `breakable: true`.
- Breakable blocks and size threshold
- New `SplitOptions`:
- `codeBlockMaxChars` (default: 2× `maxChars`) — closed blocks larger than this are treated as breakable (splittable).
- `fallbackCloseOnNestedOpen` (default: true).
- Splitting rules updated
- Paragraph/line splitting skips split points only inside non-breakable blocks; allows splits inside breakable blocks.
- Final boundary pass adjusts segment starts/ends only for non-breakable blocks; ensures monotonic, non-overlapping segments.
- Overlap start is pushed past block end only if the enclosing block is non-breakable.
- Tests
- New: `ingesters/src/utils/__tests__/RecursiveMarkdownSplitter.noStartInsideCodeBlock.test.ts` enforces that no chunk starts inside a non-breakable code block; allows starts inside breakable (malformed/oversized) blocks.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Ingestion: Markdown Splitter (TypeScript)
---…Sources:…---.chunk.meta.sourceLinkto the first URL listed in the most recent Sources block for all subsequent chunks until the next block.ChunkMetawith optionalsourceLinkand tokenizerTokenswithsourceRanges.ingesters/src/utils/__tests__/RecursiveMarkdownSplitter.sources.test.tsvalidates activeSource mapping across multiple Sources blocks.Splitter robustness and fallback for fenced code blocks
breakable: true.SplitOptions:codeBlockMaxChars(default: 2×maxChars) — closed blocks larger than this are treated as breakable (splittable).fallbackCloseOnNestedOpen(default: true).ingesters/src/utils/__tests__/RecursiveMarkdownSplitter.noStartInsideCodeBlock.test.tsenforces that no chunk starts inside a non-breakable code block; allows starts inside breakable (malformed/oversized) blocks.