165 lines
5.4 KiB
Python
165 lines
5.4 KiB
Python
"""AI summarization module.
|
||
|
||
Supports OpenAI-compatible (doubao, deepseek, openai) and Anthropic APIs.
|
||
Uses urllib.request for zero extra dependencies.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import json
|
||
import logging
|
||
import ssl
|
||
import urllib.request
|
||
from typing import Any, Dict, List
|
||
|
||
from .models import Tweet
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
SYSTEM_MESSAGE = "你是一个专业的 Twitter/X 信息流分析师，擅长提炼关键信息和发现趋势。"
|
||
|
||
|
||
def _build_prompt(tweets, language="zh-CN"):
|
||
# type: (List[Tweet], str) -> str
|
||
"""Build the summarization prompt."""
|
||
lines = []
|
||
for i, t in enumerate(tweets):
|
||
score_str = " [score: %.1f]" % t.score if t.score else ""
|
||
rt = " (RT by @%s)" % t.retweeted_by if t.is_retweet and t.retweeted_by else ""
|
||
media_str = ""
|
||
if t.media:
|
||
media_str = " [%s]" % ", ".join(m.type for m in t.media)
|
||
url_str = ""
|
||
if t.urls:
|
||
url_str = "\n Links: %s" % ", ".join(t.urls)
|
||
quoted = ""
|
||
if t.quoted_tweet:
|
||
qt = t.quoted_tweet
|
||
quoted = "\n Quoting @%s: %s..." % (qt.author.screen_name, qt.text[:100].replace("\n", " "))
|
||
|
||
text_preview = t.text.replace("\n", " ")[:300]
|
||
lines.append(
|
||
'%d. @%s (%s)%s%s\n'
|
||
' "%s"\n'
|
||
' ❤️%d 🔄%d 💬%d 🔖%d 👁️%d%s%s%s'
|
||
% (
|
||
i + 1, t.author.screen_name, t.author.name, rt, score_str,
|
||
text_preview,
|
||
t.metrics.likes, t.metrics.retweets, t.metrics.replies,
|
||
t.metrics.bookmarks, t.metrics.views,
|
||
media_str, url_str, quoted,
|
||
)
|
||
)
|
||
|
||
tweet_summaries = "\n\n".join(lines)
|
||
|
||
if language.startswith("zh"):
|
||
lang_inst = "请用中文输出。"
|
||
else:
|
||
lang_inst = "Please output in %s." % language
|
||
|
||
return (
|
||
"你是一个 Twitter/X 信息流分析师。请对以下 %d 条推文进行摘要总结。\n\n"
|
||
"要求：\n"
|
||
"1. 按主题分组（如：AI & 编程、Crypto、工具推荐、生活观点等）\n"
|
||
"2. 每组列出关键推文和核心观点，标注作者 @handle\n"
|
||
"3. 标注数据亮点（高赞/高收藏推文用 🔥 标记）\n"
|
||
"4. 最后用 2-3 句话总结今天 timeline 的整体趋势\n"
|
||
"5. %s\n\n"
|
||
"推文数据：\n\n%s"
|
||
) % (len(tweets), lang_inst, tweet_summaries)
|
||
|
||
|
||
def _call_openai(prompt, config):
|
||
# type: (str, Dict[str, Any]) -> str
|
||
"""Call OpenAI-compatible API."""
|
||
url = config.get("base_url", "").rstrip("/")
|
||
if not url.endswith("/chat/completions"):
|
||
if not url.endswith("/v1"):
|
||
url += "/v1"
|
||
url += "/chat/completions"
|
||
|
||
payload = json.dumps({
|
||
"model": config.get("model", ""),
|
||
"messages": [
|
||
{"role": "system", "content": SYSTEM_MESSAGE},
|
||
{"role": "user", "content": prompt},
|
||
],
|
||
"temperature": 0.3,
|
||
"max_tokens": 4096,
|
||
}).encode("utf-8")
|
||
|
||
req = urllib.request.Request(url, data=payload)
|
||
req.add_header("Content-Type", "application/json")
|
||
req.add_header("Authorization", "Bearer %s" % config.get("api_key", ""))
|
||
|
||
ctx = ssl.create_default_context()
|
||
with urllib.request.urlopen(req, context=ctx, timeout=120) as resp:
|
||
data = json.loads(resp.read().decode("utf-8"))
|
||
|
||
choices = data.get("choices", [])
|
||
if choices:
|
||
return choices[0].get("message", {}).get("content", "")
|
||
return ""
|
||
|
||
|
||
def _call_anthropic(prompt, config):
|
||
# type: (str, Dict[str, Any]) -> str
|
||
"""Call Anthropic Messages API."""
|
||
url = config.get("base_url", "").rstrip("/")
|
||
if not url.endswith("/messages"):
|
||
if not url.endswith("/v1"):
|
||
url += "/v1"
|
||
url += "/messages"
|
||
|
||
payload = json.dumps({
|
||
"model": config.get("model", ""),
|
||
"system": SYSTEM_MESSAGE,
|
||
"messages": [{"role": "user", "content": prompt}],
|
||
"temperature": 0.3,
|
||
"max_tokens": 4096,
|
||
}).encode("utf-8")
|
||
|
||
req = urllib.request.Request(url, data=payload)
|
||
req.add_header("Content-Type", "application/json")
|
||
req.add_header("x-api-key", config.get("api_key", ""))
|
||
req.add_header("anthropic-version", "2023-06-01")
|
||
|
||
ctx = ssl.create_default_context()
|
||
with urllib.request.urlopen(req, context=ctx, timeout=120) as resp:
|
||
data = json.loads(resp.read().decode("utf-8"))
|
||
|
||
content_blocks = data.get("content", [])
|
||
for block in content_blocks:
|
||
if block.get("type") == "text":
|
||
return block.get("text", "")
|
||
return ""
|
||
|
||
|
||
def summarize(tweets, config):
|
||
# type: (List[Tweet], Dict[str, Any]) -> str
|
||
"""Summarize tweets using the configured AI provider.
|
||
|
||
Config keys: provider, api_key, model, base_url, language
|
||
"""
|
||
api_key = config.get("api_key", "")
|
||
if not api_key:
|
||
raise RuntimeError(
|
||
"AI API key not configured.\n"
|
||
"Set ai.api_key in config.yaml or export AI_API_KEY=your_key"
|
||
)
|
||
|
||
if not tweets:
|
||
return "No tweets to summarize."
|
||
|
||
language = config.get("language", "zh-CN")
|
||
prompt = _build_prompt(tweets, language)
|
||
provider = config.get("provider", "openai")
|
||
|
||
logger.info("Calling AI (%s/%s)...", provider, config.get("model", ""))
|
||
|
||
if provider == "anthropic":
|
||
return _call_anthropic(prompt, config)
|
||
else:
|
||
return _call_openai(prompt, config)
|