---
pretty_name: funes memory
license: apache-2.0
language:
  - zh
source_datasets:
  - shareAI/ShareGPT-Chinese-English-90k
tags:
  - funes
  - agent-memory
  - agent-traces
  - embeddings
  - lance
size_categories:
  - 1K<n<10K
---

# funes memory

A [funes](https://github.com/huggingface/funes) memory: agent sessions chunked,
embedded, and written to a [Lance](https://lancedb.github.io/lance/) table — a derived
index holding verbatim passages with exact provenance, not raw transcripts.

Any agent (or you) can recall from it directly — no local index needed:

```bash
funes recall "what did we decide about …" --memory huggingface/funes-sharegpt-zh
```

Get funes:

```bash
curl -fsSL https://huggingface.co/buckets/huggingface/funes/resolve/install.sh | sh
```

## Source

2000 "Program and Code" conversations from [shareAI/ShareGPT-Chinese-English-90k](https://huggingface.co/datasets/shareAI/ShareGPT-Chinese-English-90k) (`computer_zh_26k(fixed).jsonl`, Apache-2.0), sampled with seed 144 among the conversations present in both languages with the same number of turns. Each conversation is one session, each message one turn. Indexed with stock funes (`1.6.0+dev`, commit `fe60450`).

The same conversations are in `huggingface/funes-sharegpt-zh` (Chinese), `huggingface/funes-sharegpt-en` (English) and `huggingface/funes-sharegpt-zh-e5` (Chinese, multilingual embeddings). The labeled questions over them are in [dacorvo/funes-recall-benchmark](https://huggingface.co/datasets/dacorvo/funes-recall-benchmark/tree/main/sharegpt).

<!-- funes:stats -->
| | |
|---|---|
| Chunks | 7,634 |
| Embedding model | `BAAI/bge-small-en-v1.5` |
| Updated | 2026-10-06 |
<!-- /funes:stats -->

*Published and kept current by `funes push`.*