Skip to content

gegedenice/uv-scripts

BranchesTags

Repository files navigation

uv-scripts

uv utilities : some python files remotly actionable with uv run...

OpenAlex Embedding Atlas Dashboard

A Python script that harvests works from the OpenAlex API and visualizes them using the interactive and semantic dashboard provided by the embedding-atlas library.

See README

RAG demo

Some Python utilities for building and querying RAG, see README

Idref Web services

Basic

GH_RAW="https://raw.githubusercontent.com/gegedenice/uv-scripts/main/Idref/idref-webservice-basic.py"

uv run "{GH_RAW}" --web-service idref2id --id 240229061 --json-output

Authorities alignment pipeline (persons)

GH_RAW="https://raw.githubusercontent.com/gegedenice/uv-scripts/main/Idref/idref-person-align.py"

uv run "{GH_RAW}" --name "Valérie Robert" --title "Satisfaction et vécu périopératoire des patients opérés sous anesthésie péribulbaire" --degree-type "Thèse d'exercice"

Sudoc Multiwhere web service

GH_RAW="https://raw.githubusercontent.com/gegedenice/uv-scripts/main/multiwhere-webservice.py"

uv run "{GH_RAW}" 123456789 --format table
uv run "{GH_RAW}" 123456789 --format geojson
uv run "{GH_RAW}" 123456789 --format geojson > holdings.geojson
uv run "{GH_RAW}" 123456789 --llm-mode clean_json # without llm posting
uv run "{GH_RAW}" 123456789 --llm --llm-mode summary # with llm posting
uv run "{GH_RAW}" 123456789 --llm --llm-mode summary --llm-url http://localhost:8080/v1/chat/completions --llm-model LFM2.5-1.2B-Instruct-Q8_0
uv run "{GH_RAW}" 123456789 --llm --llm-mode summary --llm-url https://albert.api.etalab.gouv.fr/v1/chat/completions --llm-model openai/gpt-oss-120b --llm-api-key sk-xxxxx

llms-openai-inference.py

GH_RAW="https://raw.githubusercontent.com/gegedenice/uv-scripts/main/llms-openai-inference.py"

List models (any provider) — raw JSON

uv run "{GH_RAW}" --provider openai --api-key "{OPENAI_API_KEY}" --list-models --json-output | jq .

Openai minimal

export OPENAI_API_KEY=sk-...
uv run "{GH_RAW}" \
  --provider openai \
  --model gpt-4o-mini \
  -u "Say hello in one sentence."

Groq minimal

export GROQ_API_KEY=grq-...
uv run "{GH_RAW}" \
  --provider groq \
  --model moonshotai/kimi-k2-instruct-0905 \
  -u "Give 3 bullet points about RAG pitfalls."

Ollama Cloud minimal

uv run "{GH_RAW}" \
  --provider ollama-cloud \
  --model kimi-k2.5:cloud \
  --api-key "{OLLAMA_API_KEY}" \
  -u "Summarize OAIS in 2 lines." \
  -s "Be concise and precise."

Albert APi minimal

uv run "{GH_RAW}" \
  --provider albert \
  --model kimi-k2.5:cloud \
  --api-key "{ALBERT_API_KEY}" \
  -u "Summarize OAIS in 2 lines." \
  -s "Be concise and precise."

Huggingface minimal

uv run "{GH_RAW}" \
  --provider huggingface \
  --hf-subpath novita/v3/openai \
  --model moonshotai/kimi-k2-instruct \
  --api-key "{hf_token}" \
  -u "{USER_PROMPT}" \
  -s "{SYSTEM_PROMPT}" \
  --temperature 0.0 \
  --max-tokens 16384

OpenAI — system prompt + decoding options + token limit

uv run "{GH_RAW}" \
  --provider openai \
  --model gpt-4o-mini \
  -u "Summarize OAIS in 2 lines." \
  -s "Be concise and precise." \
  --temperature 0.0 \
  --max-tokens 64

OpenAI — reasoning models (reasoning_effort)

uv run "{GH_RAW}" \
  --provider openai \
  --model o3-mini \
  -u "Explain the difference between SIP and AIP in OAIS." \
  --reasoning-effort medium

OpenAI — extra knobs via --options-json

uv run "{GH_RAW}" \
  --provider openai \
  --model gpt-4o-mini \
  -u "One insightful line about archival appraisal." \
  --options-json '{"presence_penalty":0.2,"frequency_penalty":0.1}'

Hugging Face Inference Router (OpenAI-compatible path)

export HF_API_KEY=hf_...
uv run "{GH_RAW}" \
  --provider hf \
  --hf-subpath openai/v1 \
  --model meta-llama/Meta-Llama-3-8B-Instruct \
  -u "What is an archival fonds?"

Self-hosted / local OpenAI-compatible server (override base URL) / Remote-hosted OpenAI-compatible server (override base URL)

export LLM_API_KEY=not-used-or-your-key
uv run "{GH_RAW}" \
  --provider openai \
  --base-url http://localhost:8000/v1 \
  --model my-local-model \
  -u "Test a local server call."

JSON output (easy to script)

uv run "{GH_RAW}" \
  --provider openai \
  --model gpt-4o-mini \
  -u "Give me a single JSON object with keys a,b." \
  --json-output | jq -r .text

Read prompt from a file

uv run "{GH_RAW}" \
  --provider openai \
  --model gpt-4o-mini \
  -u "$(cat prompt.txt)"

Script for generic MCP servers from OpenAPI specifications : mcp-server-from-openapi.py

One file, one command : Minimal MCP server generated from an OpenAPI specification using FastMCP with multiple transport features (STDIO, HTTP and Streamable-HTTP).

The server automatically fetches the OpenAPI specification and creates MCP tools based on the available endpoints.

By default, this MCP server targets the OpenAlex REST API using this provided OpenAPI specification, but it can dynamically adapt to any other API by changing the CLI parameters.

How it works

  1. Loads an OpenAPI spec from a online openapi.json file.
  2. Creates an httpx.AsyncClient pinned to the API url (the one described by the openapi.json file).
  3. Generates tools automatically with FastMCP.from_openapi(...).
  4. Starts the MCP server in the transport you choose at runtime.

Minimal launch commands

GH_RAW="https://raw.githubusercontent.com/gegedenice/uv-scripts/main/mcp-server-from-openapi.py"
API_BASE_URL="..." # Fill with your API base url
OPENAPI_SPEC_URL="..." # Fill with an OpenAPI json specification (url-based or local path)

# Default STDIO on OpenAlex API
uv run "{GH_RAW}"

# STDIO
uv run "{GH_RAW}" \
   --api-base-url "{API_BASE_URL}" --openapi-spec-url "{OPENAPI_SPEC_URL}" \
   --transport stdio

# Local HTTP
uv run "{GH_RAW}" --api-base-url "{API_BASE_URL}" --openapi-spec-url "{OPENAPI_SPEC_URL}" \
   --transport http \
   --port 3333 \
   --path /mcp

# Streamable HTTP
uv run "{GH_RAW}" --api-base-url "{API_BASE_URL}" --openapi-spec-url "{OPENAPI_SPEC_URL}" \
   --transport streamable-http \
   --port 3333 \
   --path /mcp \
   --stateless-http

Environement variables

Variable Purpose Recommended
FASTMCP_EXPERIMENTAL_ENABLE_NEW_OPENAPI_PARSER Enables FastMCP’s new OpenAPI parser Set to true before using FastMCP (already done in the script)

CLI flags

Flag Values / Type Default Notes
--api-base-url string https://api.openalex.org Base URL for the target API
--openapi-spec-url string https://smartbiblia.fr/api/openalex-openapi-server/openapi.json OpenAPI JSON URL or local file path
--transport stdio http streamable-http
--host string 0.0.0.0 HTTP modes only
--port int 3333 HTTP modes only
--path string /mcp HTTP modes only
--stateless-http flag false Required for HTTP transport modes

Configuration examples

# STDIO MCP server on your API_BASE_URL and OPENAPI_SPEC_URL variables
{
  "mcpServers": {
    "openapi-mcp-server": {
        "command": "uv",
        "args": ["run", "https://raw.githubusercontent.com/gegedenice/uv-scripts/main/mcp-server-from-openapi.py", "--api-base-url", API_BASE_URL, "--openapi-spec-url", OPENAPI_SPEC_URL],
    }     	
  }
}

#streamable-http OpenAlex MCP server
{
  "mcpServers": {
    "openapi-mcp-server": {
        "command": "uv",
        "args": ["run", "https://raw.githubusercontent.com/gegedenice/uv-scripts/main/mcp-server-from-openapi.py", "--transport", "streamable-http", "--stateless-http"],
        "url": "http://localhost:3333/mcp"
    }     	
  }
}

About

Some uv scripts utilities

Resources

Readme

License

MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages