feat: Ghost CMS integration — auto-publish blog + /publish command
Build & Deploy ResearchOwl / build-and-push (push) Successful in 6s

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
ChemaVX
2026-05-08 10:26:22 +00:00
parent 94d209dd8a
commit 83eb2359be
3 changed files with 208 additions and 2 deletions
+73
View File
@@ -153,6 +153,7 @@ async def cmd_start(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
"`/sources` — List all sources found\n"
"`/outputs` — List generated outputs\n"
"`/export` — Exportar último output como PDF\n"
"`/publish` — Publicar último output en Ghost como borrador\n"
"`/compare <tema1> vs <tema2>` — Análisis comparativo\n"
"`/costs` — Show API usage costs\n"
"`/watch <topic> [h]` — Schedule periodic research\n"
@@ -931,6 +932,77 @@ async def cmd_purge(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
await db_conn.close()
async def cmd_publish(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
if not is_authorized(update.effective_user.id):
return
chat_id = update.effective_chat.id
db_conn = await get_db()
db = ResearchDB(db_conn)
try:
from src.generator.generator import GhostPublisher, _extract_title
ghost = GhostPublisher()
if not ghost.is_configured():
await update.message.reply_text(
"❌ Ghost no configurado. Asegúrate de que `GHOST_URL` y `GHOST_API_KEY` están definidos.",
parse_mode=ParseMode.MARKDOWN
)
return
cursor = await db_conn.execute(
"SELECT * FROM research_sessions WHERE telegram_chat_id = ? ORDER BY created_at DESC LIMIT 1",
(chat_id,)
)
row = await cursor.fetchone()
if not row:
await update.message.reply_text("No hay sesiones. Usa /research primero.")
return
session = dict(row)
outputs = await db.get_outputs(session["id"])
if not outputs:
await update.message.reply_text(
"No hay outputs generados. Usa `/generate blog|report` primero.",
parse_mode=ParseMode.MARKDOWN
)
return
priority = ["blog_extended", "blog", "report_extended", "report",
"podcast_extended", "podcast", "thread"]
chosen = None
for ptype in priority:
for o in outputs:
if o["output_type"] == ptype:
chosen = o
break
if chosen:
break
if not chosen:
chosen = outputs[-1]
msg = await update.message.reply_text("📤 Publicando en Ghost como borrador…")
title = _extract_title(chosen["content"]) or session["topic"]
result = await ghost.publish_draft(title, chosen["content"])
post = result["posts"][0]
admin_url = f"{ghost.url}/ghost/#/editor/post/{post['id']}"
await msg.edit_text(
f"✅ *Publicado en Ghost como borrador*\n\n"
f"📝 Título: `{title}`\n"
f"🔗 Editar: {admin_url}",
parse_mode=ParseMode.MARKDOWN
)
except Exception as e:
logger.error("Publish to Ghost failed", error=str(e))
await update.message.reply_text(f"❌ Error publicando en Ghost: {str(e)[:200]}")
finally:
await db_conn.close()
async def cmd_compare(update: Update, ctx: ContextTypes.DEFAULT_TYPE):
if not is_authorized(update.effective_user.id):
return
@@ -1097,6 +1169,7 @@ def create_bot() -> Application:
app.add_handler(CommandHandler("process", cmd_process))
app.add_handler(CommandHandler("cancel", cmd_cancel))
app.add_handler(CommandHandler("purge", cmd_purge))
app.add_handler(CommandHandler("publish", cmd_publish))
app.add_handler(CommandHandler("compare", cmd_compare))
return app
+4
View File
@@ -33,6 +33,10 @@ class Settings(BaseSettings):
chunk_overlap: int = Field(100, env="CHUNK_OVERLAP")
quality_threshold: float = Field(0.3, env="QUALITY_THRESHOLD") # 0-1, chunks below discarded
# Ghost CMS
ghost_url: Optional[str] = Field(None, env="GHOST_URL")
ghost_api_key: Optional[str] = Field(None, env="GHOST_API_KEY")
# Alerts
cost_alert_threshold: float = Field(0.15, env="COST_ALERT_THRESHOLD")
+131 -2
View File
@@ -2,6 +2,12 @@
ResearchOwl Generators
Produces structured outputs from processed research using Claude or Ollama
"""
import base64
import hashlib
import hmac
import json
import time
import structlog
from src.config import settings
@@ -196,6 +202,92 @@ Material disponible (resumen):
"""
# ─── Ghost CMS ────────────────────────────────────────────────────────────────
def _b64url(data: bytes | str) -> str:
if isinstance(data, str):
data = data.encode()
return base64.urlsafe_b64encode(data).rstrip(b"=").decode()
def _extract_title(content: str) -> str:
"""Return first H1 heading from markdown, skipping the ResearchOwl header block."""
in_header = False
for line in content.splitlines():
stripped = line.strip()
if stripped == "---":
in_header = not in_header
continue
if in_header:
continue
if stripped.startswith("# ") and not stripped.startswith("## "):
return stripped[2:].strip()
return ""
def _strip_researchowl_header(content: str) -> str:
"""Remove the ---...--- metadata block that ResearchOwl prepends to outputs."""
lines = content.splitlines(keepends=True)
dashes_seen = 0
for i, line in enumerate(lines):
if line.strip() == "---":
dashes_seen += 1
if dashes_seen == 2:
return "".join(lines[i + 1:]).lstrip("\n")
return content
class GhostPublisher:
def __init__(self):
self.url = (settings.ghost_url or "").rstrip("/")
self.api_key = settings.ghost_api_key or ""
def is_configured(self) -> bool:
return bool(self.url and self.api_key)
def _make_token(self) -> str:
key_id, secret = self.api_key.split(":", 1)
now = int(time.time())
header = _b64url(json.dumps({"alg": "HS256", "typ": "JWT", "kid": key_id}))
payload = _b64url(json.dumps({"iat": now, "exp": now + 300, "aud": "/admin/"}))
signing = f"{header}.{payload}"
sig = _b64url(
hmac.new(bytes.fromhex(secret), signing.encode(), hashlib.sha256).digest()
)
return f"{signing}.{sig}"
async def publish_draft(self, title: str, markdown_content: str,
tags: list[str] | None = None) -> dict:
import aiohttp as _aio
import markdown as _md
clean = _strip_researchowl_header(markdown_content)
html = _md.markdown(clean, extensions=["extra"])
token = self._make_token()
body = {
"posts": [{
"title": title,
"html": html,
"status": "draft",
"tags": [{"name": t} for t in (tags or ["investigacion"])],
}]
}
async with _aio.ClientSession() as sess:
async with sess.post(
f"{self.url}/ghost/api/admin/posts/",
json=body,
headers={
"Authorization": f"Ghost {token}",
"Accept-Version": "v5.0",
},
) as resp:
if resp.status not in (200, 201):
text = await resp.text()
raise ValueError(f"Ghost API {resp.status}: {text[:300]}")
return await resp.json()
# ─── Output generation ────────────────────────────────────────────────────────
class OutputGenerator:
def __init__(self, db: ResearchDB, ollama: OllamaClient, processor: ContentProcessor):
self.db = db
@@ -250,8 +342,26 @@ class OutputGenerator:
# Save to DB
await self.db.save_output(session_id, output_type, full_output)
# Auto-publish to Ghost for blog outputs
ghost_notice = ""
if output_type in (OutputType.BLOG, OutputType.BLOG_EXTENDED):
ghost = GhostPublisher()
if ghost.is_configured():
try:
title = _extract_title(full_output) or topic
result = await ghost.publish_draft(title, full_output)
post = result["posts"][0]
ghost_notice = (
f"\n\n---\n"
f"📤 *Borrador publicado en Ghost*\n"
f"Editar: {ghost.url}/ghost/#/editor/post/{post['id']}"
)
logger.info("Auto-published blog to Ghost", post_id=post["id"])
except Exception as e:
logger.warning("Auto-publish to Ghost failed", error=str(e))
logger.info("Output generated", type=output_type, length=len(full_output))
return full_output
return full_output + ghost_notice
async def _generate(self, prompt: str, system: str, output_type: OutputType,
session_id: int | None = None) -> str:
@@ -403,9 +513,28 @@ class OutputGenerator:
full_output = header + "\n\n" + full_content
await self.db.save_output(session_id, output_type, full_output)
# Auto-publish to Ghost for extended blog outputs
ghost_notice = ""
if output_type == OutputType.BLOG_EXTENDED:
ghost = GhostPublisher()
if ghost.is_configured():
try:
title = _extract_title(full_output) or topic
result = await ghost.publish_draft(title, full_output)
post = result["posts"][0]
ghost_notice = (
f"\n\n---\n"
f"📤 *Borrador publicado en Ghost*\n"
f"Editar: {ghost.url}/ghost/#/editor/post/{post['id']}"
)
logger.info("Auto-published extended blog to Ghost", post_id=post["id"])
except Exception as e:
logger.warning("Auto-publish to Ghost failed (extended)", error=str(e))
logger.info("Extended output generated", type=output_type,
sections=len(sections), length=len(full_output))
return full_output
return full_output + ghost_notice
async def _generate_raw(self, prompt: str,
session_id: int | None = None) -> str: