---
id: "0109m"
title: "issues_api service: HTTP backend para issues+flows (skill_tree, kanban, dashboards)"
status: pendiente
type: feature
domain:
  - meta
  - apps-infra
scope: app-scoped
priority: media
depends: []
blocks:
  - "0109h2"
related:
  - "0109"
  - "0106"
created: 2026-05-18
updated: 2026-05-18
tags:
  - service
  - go
  - http
  - issues
  - flows
  - api
---

# 0109m — issues_api service

## Por que

`skill_tree.exe` corre en Windows nativo. Necesita leer `dev/issues/*.md` + `dev/flows/*.md` que viven en WSL `~/fn_registry/`. Opciones:

| Solucion | Pro | Contra |
|---|---|---|
| UNC `\\wsl.localhost\Ubuntu\home\lucas\fn_registry` | Cero infra extra | Hardcodea distro/user. No funciona si WSL caido. |
| Service HTTP (este issue) | Robusto, reusable por otras apps (kanban, web dashboard) | Mas piezas que mantener |
| Embeber YAML parser + nested fields | Self-contained | Mucha logica duplicada |

Fix UNC ya esta hecho (0109l). Pero el patron canonico del registry son services (services_api, sqlite_api, registry_api). Este issue formaliza `issues_api`.

## Spec

Service Go HTTP en puerto **8486**. Patron identico a `services_api` (issue 0106).

### Endpoints

- `GET /api/health` → `{"status":"ok"}`
- `GET /api/issues` → array JSON con TODOS los issues (open + completed)
- `GET /api/issues/:id` → un issue concreto con body markdown completo
- `GET /api/flows` → array JSON con todos los flows
- `GET /api/flows/:id` → un flow concreto
- `GET /api/stats` → counts por status/domain/type (agregados para Dashboard)
- `POST /api/issues` → crea nuevo issue (escribe `.md`). Body: `{id, title, type, domain, depends, body, dod}`. Devuelve `{path}`.
- `POST /api/flows` → crea nuevo flow. Mismo shape adaptado a frontmatter de flows.

### Schema JSON issue

```json
{
  "id": "0109",
  "title": "...",
  "status": "pendiente",
  "status_eff": "pendiente_unlocked",
  "type": "epic",
  "domain": ["meta", "cpp-stack"],
  "priority": "media",
  "depends": [],
  "blocks": [],
  "related": ["0085"],
  "tags": ["skill-tree"],
  "created": "2026-05-17",
  "updated": "2026-05-18",
  "file_path": "dev/issues/0109-skill-tree-app-roadmap.md",
  "body_md": "(opcional, solo en /api/issues/:id)",
  "dod": [{"text":"App existe","done":true}, ...]
}
```

### Implementacion

```go
// apps/issues_api/main.go
package main

import (
    "encoding/json"
    "net/http"
    "log"
    "os"
    // reusa funciones del registry:
    // - extract_frontmatter (Go port; existe extract_frontmatter_py_core, pero
    //   en C++ existe parse_md_frontmatter_cpp_core — necesitamos Go port o
    //   reusar yaml.v3 directo aqui).
)

func main() {
    root := os.Getenv("FN_REGISTRY_ROOT")
    if root == "" { log.Fatal("FN_REGISTRY_ROOT not set") }
    mux := http.NewServeMux()
    mux.HandleFunc("/api/health", health)
    mux.HandleFunc("/api/issues", listIssues(root))
    mux.HandleFunc("/api/issues/", showIssue(root))
    mux.HandleFunc("/api/flows", listFlows(root))
    mux.HandleFunc("/api/flows/", showFlow(root))
    mux.HandleFunc("/api/stats", stats(root))
    log.Fatal(http.ListenAndServe(":8486", mux))
}
```

Frontmatter: existe `extract_frontmatter_py_core` (Python). Hace falta:
- **Opcion A**: crear `parse_md_frontmatter_go_core` (port del C++). Reusable por otros services Go.
- **Opcion B**: usar `gopkg.in/yaml.v3` directo dentro de `issues_api`.

Recomiendo **A** — Go port reusable. Pattern: header `parse_md_frontmatter.go` en `functions/core/`.

### Frontmatter service: block en app.md

```yaml
service:
  port: 8486
  health_endpoint: /api/health
  health_timeout_s: 3
  systemd_unit: issues_api.service
  systemd_scope: user
  restart_policy: always
  runtime: systemd-user
  pc_targets:
    - aurgi-pc
    - home-wsl
  is_local_only: false
```

### Skill_tree consume

Cambio en `skill_tree/main.cpp`:
- Si `discover_registry_root()` falla, intentar HTTP `http://localhost:8486/api/issues`.
- Si responde 200 → parsear JSON y poblar `g_scan.nodes`.
- Fallback a UNC + file scan si no.

Cliente HTTP en C++:
- Linux: `libcurl` (ya disponible) o `popen("curl ...")`.
- Windows: WinHTTP nativo o curl.exe (Windows 10+ trae curl).
- Mas simple: spawn `curl.exe -s http://...` y parsear stdout (nlohmann json — habria que vendor).

## Sub-issues

- **0109m1** — crear `parse_md_frontmatter_go_core` port del C++ (mismas semanticas, mismos tests).
- **0109m2** — scaffolder `issues_api` app + service block + systemd unit.
- **0109m3** — implementar endpoints GET (list, show, stats).
- **0109m4** — implementar POST issues + POST flows (escribir .md valido).
- **0109m5** — cliente HTTP en skill_tree + fallback chain (env > walk > UNC > HTTP).

## Beneficios cross-app

Una vez issues_api existe, reutilizan:
- `kanban` (issue 0058 sync) podria leer/sync issues.
- Frontend web dashboard ya tendria backend.
- CI/automation puede crear issues via POST en lugar de editar .md a mano.
- `services_monitor` mostrara `issues_api` como service mas (auto via `tag: service`).

## DoD

- [ ] `apps/issues_api/` scaffoldada via `/cpp-app` ?? no — Go service via `/app`.
- [ ] Endpoints listed devuelven JSON valido.
- [ ] systemd unit instalado (`issues_api.service`).
- [ ] Visible en `services_monitor` como service activo.
- [ ] skill_tree consume HTTP cuando esta disponible, cae a UNC/file si no.
- [ ] Tests Go basicos (list, show, post).