README.md

# AgentPulse — Data and Code Release

**NeurIPS 2026 Datasets & Benchmarks Track companion artifact.**

This bundle contains all data, code, and reproduction scripts for the
AgentPulse evaluation framework described in the paper
*AgentPulse: A Continuous Multi-Signal Framework for Evaluating AI Agents in Deployment*.

License: **CC BY 4.0** (see `LICENSE`).

---

## Contents

```
AgentPulse-Data-v6/
├── README.md                  ← this file
├── LICENSE                    ← CC BY 4.0
├── croissant.json             ← Croissant metadata (JSON-LD)
├── data/
│   ├── csv/                   ← 18 CSV exports of every relevant DB table
│   └── sqlite/                ← (optional) full SQLite database, if present
├── code/
│   ├── collectors/            ← 18 signal-collector implementations
│   ├── scoring/               ← NLP pipeline + factor composite + data quality
│   ├── pipeline/              ← scheduler / orchestration
│   ├── config.py              ← global pipeline configuration
│   ├── main.py                ← entrypoint (`python main.py --serve`, etc.)
│   └── requirements.txt       ← Python dependencies
├── paper/
│   ├── neurips_agentpulse.tex ← LaTeX source
│   ├── neurips_agentpulse.pdf ← compiled paper
│   ├── neurips_2026.sty       ← NeurIPS 2026 style file
│   ├── checklist_.tex         ← NeurIPS Paper Checklist
│   ├── regen_figures.py       ← regenerates all 5 paper figures from the data
│   └── fig*.{png,pdf}         ← compiled figures
└── docs/
    └── schema.md              ← per-table column documentation
```

---

## Data tables (`data/csv/`)

| File                              | Rows    | Description |
|-----------------------------------|---------|-------------|
| `agent_scores.csv`                | 13,504  | Composite + per-factor scores per agent per category, time-stamped. The headline scoring output. |
| `agent_signals_raw.csv`           | 3,283   | Raw per-signal observations per agent (pre-aggregation). |
| `agent_benchmark_signals.csv`     | 1,563   | Published benchmark scores (SWE-bench, GAIA, WebArena, HumanEval+, τ-bench). |
| `agent_github_history.csv`        | 62      | GitHub stars / contributors / commits time-series. |
| `agent_pypi_history.csv`          | 1,440   | PyPI / npm download history. |
| `sentiment_scores.csv`            | 60,639  | NLP-scored texts (VADER + TextBlob + FinBERT + DistilBERT-SST2 ensemble). |
| `data_quality.csv`                | 38,965  | Composite quality score per text (uniqueness × bot × credibility × specificity). |
| `models.csv`                      | 148     | Agent + model registry with vendor / category metadata. |
| `bluesky_signals.csv`             | 35,326  | Per-platform raw text + engagement. |
| `reddit_signals.csv`              | 2,413   | "" |
| `hn_signals.csv`                  | 21,205  | "" |
| `stackoverflow_signals.csv`       | 571     | "" |
| `github_signals.csv`              | 23,288  | "" |
| `github_discussions_signals.csv`  | 2,090   | "" |
| `devto_signals.csv`               | 596     | "" |
| `mastodon_signals.csv`            | 1,160   | "" |
| `v2ex_signals.csv`                | 1,794   | "" |
| `devtools_signals.csv`            | 89      | VS Code / IDE marketplace install counts. |

All CSVs are UTF-8 encoded with a header row. Timestamps are ISO-8601
in UTC. See `docs/schema.md` for full column documentation.

---

## Reproducing paper results

```bash
cd code/
pip install -r requirements.txt

# Reproduce all figures (writes into ../paper/)
python ../paper/regen_figures.py

# Re-derive composite scores from raw signals
python -m scoring.agent_scoring_v2 --rebuild

# Re-run the full pipeline end-to-end (continuous mode)
python main.py
```

The cross-factor predictive validity result (Table 3, ρₛ=0.52, p<0.01,
n=35) is computed in `code/scoring/agent_scoring_v2.py` and can be
re-derived directly from `data/csv/agent_scores.csv` and external
adoption proxies.

---

## Provenance

All signals are collected from public APIs with documented rate-limits
under each API's terms of service. Specifically:

- **GitHub**: REST + GraphQL APIs (60 req/hr unauth, 5,000 req/hr authed)
- **PyPI**: BigQuery public dataset
- **VS Code Marketplace**: Public extension API
- **Bluesky / Mastodon**: AT-Protocol / ActivityPub public firehose
- **Reddit / Hacker News / Stack Overflow / Lemmy / Lobsters / Dev.to / V2EX**:
  Documented public APIs
- **Benchmark scores**: Scraped from the official leaderboard pages
  cited in the paper (SWE-bench, GAIA, WebArena, HumanEval+, τ-bench)

No private data, no scraped paywalled content, no PII beyond public
authorship metadata is included. The release contains no individual
user records — only aggregated per-text metadata required for sentiment
scoring (post id, author handle if public, engagement counts).

---

## Citation

```
@inproceedings{agentpulse2026,
  title = {AgentPulse: A Continuous Multi-Signal Framework for Evaluating
           AI Agents in Deployment},
  author = {Anonymous},
  booktitle = {NeurIPS Datasets and Benchmarks Track},
  year = {2026}
}
```

## Contact

Per the NeurIPS double-blind review policy, author contact is withheld
during review. Issues with the artifact may be reported via the
submission system.