Ask questions about data too large for a context window.
Droste is an open source Recursive Language Model (RLM) engine. It gives a model a Python REPL over your data, so the model can search, filter, count, and delegate small pieces of interpretation without loading the full corpus into its context.
uvx droste "which customer had a failed charge, and why?" server.log
uvx droste "which plan has the highest refund rate vs its MRR?" shop.db
uvx droste "how do the authentication flows differ?" ./docs
Code handles work that should be exact, such as searching, counting, joining, and aggregating. Model calls handle the parts that require interpretation. Droste combines both while keeping the full dataset outside the model's context window.
The root model sees only what its code prints and the bounded results of any subcalls it makes. Runs have explicit limits for iterations, subcalls, time, and model output.
Install the command and sign in once:
uv tool install droste
droste loginThen ask a question about files, directories, SQLite databases, or piped text:
droste "what changed between these?" report.txt logs.txt
droste "which customers churned last month?" app.db
tail -5000 app.log | droste "why did it crash?"For a one-off run, use uvx droste instead of installing it. The command also
supports an existing ModelRelay key or your own OpenAI-compatible or Anthropic
endpoint. See the quickstart for credentials, directories,
SQLite, and useful flags.
- Exact counts, aggregates, and joins over logs, exports, or databases.
- Classification or review across many records using bounded model batches.
- Questions that need both computation and close reading.
If the data already fits comfortably in a context window, or the task is open-ended agent work rather than a question about a corpus, use a general agent instead. The quickstart includes examples for logs, archives, and SQLite.
Published runs show the largest gains on tasks that require aggregation across scattered or oversized context.
| Benchmark | Direct baseline | Droste | Recorded cost |
|---|---|---|---|
| OOLONG, 131K tokens | 0.6020 | 0.6432 | $10.16 vs $26.18 |
| OOLONG-Pairs, 32K tokens | 0.034 | 0.80 | $2.14 vs $2.50 |
| BrowseComp-Plus, 6.0M to 11.1M tokens | Could not run | 0.9400 | $24.54 |
These are individual published runs, not universal quality or cost guarantees. See the results and caveats, or open the repository's benchmark guide for artifacts and reproduction steps.
The same package is a dependency-free Python library. You can supply your own models and data providers, set hard compute budgets, and consume structured traces. Start with the embedding guide. Protocol and host implementers can use the technical reference.
uv sync
uv run pytest
uv buildSee CONTRIBUTING.md for the contribution workflow.
Apache-2.0. See LICENSE.