Document search app
Ingest a folder of documents, chunk them, and answer searches with ranked, linked passages.
Brief ready; automated checks coming soonIntermediateAbout 6 hoursRAG engineering
What you'll build
- Ingest Markdown or text files and keep a link from each chunk to its source
- Support keyword and embedding-based search
- Show the top results with highlighted matches and source links
- Measure recall@5 on at least ten questions with known answers
Milestones
Work through these in order. Each one ends with a check you can verify yourself.
- 1Milestone 1: Ingest and chunkSplit documents into chunks with source and position metadata.Done when: Every chunk links back to its file and line range.
- 2Milestone 2: Keyword searchRank chunks with tf-idf or BM25.Done when: Exact-term questions return the right chunk in the top 3.
- 3Milestone 3: Vector searchEmbed chunks and rank by cosine similarity.Done when: Paraphrased questions return the right chunk in the top 5.
- 4Milestone 4: EvaluateCompare both methods on a labeled question set.Done when: A table of recall@5 for keyword and vector search.
Test criteria
What a finished project must demonstrate.
- Recall@5 is computed from labeled question–chunk pairs
- Each result links to the exact source location
- Empty or unknown queries return a clear empty state
An automated runner for these criteria is planned. Until it ships, check each one yourself; nothing here is graded for you yet.
Rubric
Rate your work against each criterion. Exemplary isn't required; be honest about where you are.
Reflection
A few minutes of reflection makes the lessons stick.