Nathan9/dump / datasets /README.md
Nathan9's picture
|
download
raw
1.65 kB
metadata
pretty_name: Funes recall store (Kimi K2.6 Claude Code traces)
source_datasets:
  - armand0e/kimi-k2.6-claude-code-traces
language:
  - en
tags:
  - funes
  - recall
  - embeddings
  - agent-traces
  - lance
size_categories:
  - 10K<n<100K

Funes recall store — Kimi K2.6 Claude Code traces

A funes recall store built by indexing armand0e/kimi-k2.6-claude-code-traces (36 Claude Code-style coding-agent sessions generated with Kimi K2.6).

What this is

This is not a raw trace dataset — it is a pre-built funes index: the source sessions chunked into content blocks and embedded, stored as a Lance table (chunks.lance).

Source armand0e/kimi-k2.6-claude-code-traces (36 sessions)
Chunks 10,348
Embedding model BAAI/bge-small-en-v1.5 (384-dim)
Format Lance

Each row is one content-block chunk with its embedding and provenance (session_id, turn_uuid, role, block_type, tool_name, source_path, harness, timestamps). Secrets were redacted at index time.

Usage

funes recall "your query" --store dacorvo/funes-armand0e-kimi-k2.6-claude-code-traces

License and attribution

This store is derived from armand0e/kimi-k2.6-claude-code-traces and redistributes verbatim excerpts of it (the chunk text column). At the time of indexing the upstream dataset declared no license, so no license is asserted here; any terms of the upstream dataset apply. Attribution to the upstream dataset is required.

Xet Storage Details

Size:
1.65 kB
·
Xet hash:
5d56b232108208e5520292c0d7c51019532ff257283cab443c3ec95445063ca4

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.