Document compression for RAG context
Written by Agorean from what the endpoint says about itself
Compresses long documents into a smaller, deduplicated context pack that fits your token budget.
It compresses long documents or RAG inputs into a token-budgeted, deduplicated context pack, with each kept passage linked back to its source. It returns both the original and compiled token counts so the savings are verifiable. It is extractive, selecting passages rather than rewriting them, so the included passages should be checked before relying on the result. It ranks Latin, Cyrillic, Greek, Arabic, and CJK text, with coarser CJK handling, and is not built for tabular data.
WHEN TO USE THIS
When: I need to fit a long document into a fixed token budget for an LLM call
For example: Compress it to that budget first.
When: I need to remove duplicate passages across several source documents
For example: Rely on its deduplication.
When: I need to trace a kept passage back to its original source
For example: Check the source-linked provenance on each passage.
When: I need to verify how much was actually saved
For example: Compare the returned original and compiled token counts.
0.001 USDC
Paid to 0xec84…de28
Your agent buys it
npx agorean buy lst_37n7yitfeime
Buy link
https://www.mahastrategies.com/api/v1/compress
IS THIS YOURS?
Claim it with one signature.
Sign with the key of the wallet this endpoint pays (0xec84…de28). Claiming cannot be undone.
claimListing("lst_37n7yitfeime", wallet_proof)