novita-social-monitor（诺维塔社会监测）

Novita Social Monitor

AI industry intelligence gathering through sequential X/Twitter account review.

Output Language

Default: Chinese (中文) - All reports, summaries, and records should be in Chinese unless the user explicitly requests English.

Commands

Get Following List

twitterapi user following Jax_Zhang_4R --limit 200

Response includes: id, screen_name, name, description, followers_count, verified for each account. No additional API calls needed for account info.

Fetch Tweets (Time-Filtered)

# Read last check time from state file
STATE_FILE="$HOME/.novita-monitor/state.json"
LAST_CHECK=$(cat "$STATE_FILE" | grep -o '"last_check": "[^"]*"' | cut -d'"' -f4 2>/dev/null || echo "2026-01-01_00:00:00_UTC")

# Get current time
UNTIL=$(date -u +"%Y-%m-%d_%H:%M:%S_UTC")

# Fetch tweets from last check to now
twitterapi tweet search "from:<account> since:${LAST_CHECK} until:${UNTIL}" --limit 3 --compact

# Update state file with current time
echo "{\"last_check\": \"$UNTIL\"}" > "$STATE_FILE"

Fallback (No Time Filter)

twitterapi user tweets <account> --limit 3 --compact

Search Brand Mentions

twitterapi tweet search "@novita_labs" --limit 50 --compact

Processing Workflow

For each account in the following list:

1. 按描述分类 - 使用 following 列表返回的 description 字段确定分类: 注意: following 命令已返回 description、name、followers_count 和 verified 状态 - 无需单独获取账户信息。

- LLM 竞争对手: LLM 提供商、模型托管、推理 API - GPU 竞争对手: GPU 云、计算基础设施 - 合作伙伴: 集成合作伙伴、生态协作者 - 生态领袖: 主要平台 (OpenAI、Anthropic 等) - 中国开源: 中国 AI 模型提供商 - AI KOL: 行业思想领袖、研究人员 - 未分类: 一般 AI 行业相关性

1. 获取推文 - 使用上述时间过滤命令
2. 按分类分析:

- 竞争对手: 产品发布、定价、合作 - 合作伙伴: 集成更新 - 生态: 平台变化 - 检查互动: 100-500 (有趣)、500-2000 (值得关注)、2000+ (重要)

1. 立即记录 - 重要: 在完成每个账户审查后立即保存结果到文件，然后再处理下一个账户。这确保:

- 如果进程中断不会丢失数据 - 可以进行增量汇总 - 为将来提供审计追踪

Record Format & File Storage

Directory Structure

~/.novita-monitor/
├── <datetime>/                # Per-run directory (e.g., 2026-02-26_14-30)
│   ├── @<account>.md          # Individual account reviews
│   ├── brand-mentions.md      # @novita_labs brand mentions
│   └── summary.md             # End-of-run summary
└── state.json                 # Last check timestamp

DateTime format: YYYY-MM-DD_HH-MM (UTC)

Each execution creates a new directory based on the start time:

# Get current datetime for directory name
RUN_DATETIME=$(date -u +"%Y-%m-%d_%H-%M")
mkdir -p "$HOME/.novita-monitor/$RUN_DATETIME"

Individual Account Record

File: ~/.novita-monitor/<datetime>/@<account>.md

Create immediately after reviewing each account.

# @<account> - <日期>

## 账户信息
- **分类**: [分类]
- **粉丝数**: [数量]
- **认证**: [是/否]

## 推文分析

### [摘要]
- **互动**: [点赞] 点赞, [转发] 转发
- **时间**: [时间戳]
- **链接**: https://x.com/<account>/status/<tweet_id>
- **相关性**: [为什么重要]

每条推文必须包含原始链接。

Brand Mentions Record

File: ~/.novita-monitor/<datetime>/brand-mentions.md

# 品牌提及 - <日期时间>

## 概述
总提及数: [数量]

## 重要提及
### @<提及账户>
- **推文**: [摘要]
- **链接**: https://x.com/<account>/status/<tweet_id>
- **情感**: [正面/负面/中性]
- **需要行动**: [是/否]

End-of-Run Summary

File: ~/.novita-monitor/<datetime>/summary.md

# 运行总结 - <日期时间>

## 已审查账户
总计: [数量]

## 按分类统计
- **LLM 竞争对手**: [数量] - [关键洞察]
- **GPU 竞争对手**: [数量] - [关键洞察]
- **合作伙伴**: [数量] - [关键洞察]
- **生态领袖**: [数量] - [关键洞察]
- **中国开源**: [数量] - [关键洞察]
- **AI KOL**: [数量] - [关键洞察]
- **未分类**: [数量]

## 重大动态
[需要关注的关键事件]

## 行动项
- [ ] [行动 1]
- [ ] [行动 2]

Data Aggregation (Optional)

To aggregate data across multiple runs:

# Get all summaries from today
cat ~/.novita-monitor/2026-02-26_*/summary.md

# Get all records for a specific account across all runs
find ~/.novita-monitor -name "@github.md" -exec cat {} \;

State File

Location: ~/.novita-monitor/state.json

{
  "last_check": "2026-02-26_14:50:00_UTC"
}

Purpose: Records last execution time. Each run reads this time as since: value, then updates to current time after fetching tweets.

First run: Create file manually or script will use default 2026-01-01_00:00:00_UTC.

Language Preference

Default: All reports are generated in Chinese (中文).

English output: If you explicitly request English reports, specify at the start of your request:

"请生成英文报告" or "Generate reports in English"

This applies to:

• Individual account records
• Brand mentions summaries
• End-of-run summaries
• All analysis and insights