Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/run-llama/parsebench/integrate-pipelinegit clone --depth 1 https://github.com/run-llama/ParseBenchWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/run-llama/parsebench/integrate-pipeline)<a href="https://agentmods.dev/commands/run-llama/parsebench/integrate-pipeline"><img src="https://agentmods.dev/badge/commands/run-llama/parsebench/integrate-pipeline.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.01745 |
| Opus 5 | $0.00000 | $0.00873 |
| Sonnet 5 | $0.00000 | $0.00349 |
| Haiku 4.5 | $0.00000 | $0.00175 |
Grade A, and why
integrate-pipeline scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 175 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Integrate a new document parsing pipeline into ParseBench: $ARGUMENTS
You are integrating a new pipeline into the ParseBench benchmark. The user will provide a pipeline name and any relevant context (API docs, SDK links, product website, etc.). Your job is to create all the files needed so that uv run parse-bench run <pipeline_name> works end-to-end.
Step 1: Understand the provider
Before writing any code, research the provider:
- If the user gave a URL, fetch and read it to understand the API/SDK.
- Determine:
- Product type: Is this a
PARSEprovider (PDF -> markdown) orLAYOUT_DETECTIONprovider (PDF -> bounding boxes)? - Integration style: Cloud API (needs API key), self-hosted model (needs endpoint URL), or local library (no external deps)?
- SDK/API pattern: Does it have a Python SDK? REST API? What's the auth method?
- Input format: Does it accept PDF files directly, or does it need images (page screenshots)?
- Output format: What does the raw response look like? Markdown? HTML? JSON with pages?
- Product type: Is this a
Step 2: Find the closest existing provider to use as a template
Look at the existing providers and pick the best template:
- Cloud API with Python SDK (e.g., OpenAI, Anthropic, Google): Copy from
src/parse_bench/inference/providers/parse/openai.pyoranthropic_haiku.py - Cloud API with REST calls: Copy from
src/parse_bench/inference/providers/parse/reducto.pyorchunkr.py - Self-hosted vLLM endpoint: Copy from
src/parse_bench/inference/providers/parse/gemma4.pyorqwen3_5.py - Local library (no API): Copy from
src/parse_bench/inference/providers/parse/pymupdf.pyortesseract.py - Layout detection: Copy from
src/parse_bench/inference/providers/layoutdet/docling.py
Read the template file to understand the exact pattern.
Step 3: Create the provider file
Create src/parse_bench/inference/providers/parse/<provider_name>.py (or layoutdet/ for layout detection).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 175 lines · 0 tokens per session scan A a432856d997e
integrate-pipeline is a command published in the GitHub repository run-llama/ParseBench (559 stars, last pushed today), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,745 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
output
Generate output artifacts from active wiki content — summaries, reports, study guides, slide outlines, timelines, glossaries, comparisons. Outputs are filed back into the wiki.
gen-plan
Generate implementation plan from draft document.
daemon.de
Hintergrunddienste über macOS launchd LaunchAgents verwalten.
daemon.pt
Execute o asiai como daemon em segundo plano no Mac: monitoramento auto-start, dashboard web e métricas Prometheus no boot.
detect.de
Automatische Erkennung laufender LLM-Inferenz-Engines auf Ihrem Mac. 3-Schichten-Kaskade — Konfiguration, Port-Scan, Prozesserkennung.
doctor.pt
Diagnostique problemas de inferência LLM no Mac: asiai doctor verifica saúde dos motores, conflitos de porta, carregamento de modelos e status da GPU.