---
name: init_jupyter_analysis
kind: pipeline
lang: bash
domain: pipelines
version: "1.0.0"
purity: impure
signature: "init_jupyter_analysis(nombre: string, [...paquetes_extra: string]) -> void"
description: "Inicializa un analisis Jupyter completo en analysis/{nombre}/ con venv, paquetes, launcher, MCP y reglas para agentes Claude. Acepta paquetes extra opcionales."
tags: [jupyter, analysis, setup, pipeline, bash, launcher]
uses_functions:
  - assert_command_exists_bash_shell
  - find_free_port_bash_shell
  - init_uv_venv_bash_infra
  - uv_add_packages_bash_infra
  - write_jupyter_launcher_bash_infra
  - write_mcp_jupyter_config_bash_infra
  - write_claude_jupyter_rules_bash_infra
  - write_jupyter_registry_kernel_bash_infra
uses_types: []
returns: []
returns_optional: false
error_type: "error_go_core"
imports: []
tested: false
tests: []
test_file_path: ""
file_path: "bash/functions/pipelines/init_jupyter_analysis.sh"
---

## Ejemplo

```bash
# Analisis basico
./init_jupyter_analysis.sh finanzas

# Con paquetes extra
./init_jupyter_analysis.sh duckdb polars duckdb
./init_jupyter_analysis.sh ml scikit-learn torch

# Via fn run
fn run init_jupyter_analysis finanzas
fn run init_jupyter_analysis ml scikit-learn torch
```

## Flujo

1. `assert_command_exists` — verifica que uv o python3 estan disponibles
2. Crea estructura `analysis/{nombre}/notebooks/` y `analysis/{nombre}/data/`
3. `init_uv_venv` — crea venv en `analysis/{nombre}/.venv/`
4. `uv_add_packages` — instala jupyter, jupyterlab, jupyter-collaboration, jupyter-mcp-server, pandas, numpy, matplotlib + extras
5. `write_jupyter_launcher` — genera `run-jupyter-lab.sh` con modo colaborativo
6. `find_free_port` + `write_mcp_jupyter_config` — detecta puerto libre y genera `.mcp.json`
7. `write_claude_jupyter_rules` — genera `.claude/CLAUDE.md` con reglas de agente
8. `write_jupyter_registry_kernel` — genera IPython startup con `fn_query`, `fn_search`, `fn_code` y acceso a `python/functions/`

## Notas

Cada analisis es independiente (propio venv, propio Jupyter, propio MCP). Mismo patron que `apps/` pero para exploraciones no reutilizables.

El pipeline usa `set -euo pipefail` — cualquier fallo detiene la ejecucion.

Paquetes base siempre incluidos: jupyter, jupyterlab, jupyter-collaboration, jupyter-mcp-server, pandas, numpy, matplotlib. Los paquetes extra se añaden a estos.