LinkedIn & X Auto Poster (Lambda)

Fetches tech/AI news from Reddit and X, generates one professional LinkedIn post and one concise X post using OpenAI (gpt-4o), posts them, and logs status to MongoDB. Designed to run as an AWS Lambda on a schedule via EventBridge.

What it does

• Pulls items from:
  • Reddit (multiple tech/programming subreddits, keyword filtered)
  • X Recent Search (robust keyword trimming and fallback; rate-limit aware)
• Uses OpenAI to select the single most valuable item and generate:
  • LinkedIn post (professional, with Source: link and spaced hashtags)
  • X post (<= 280 chars + hashtags)
• Posts to LinkedIn (UGC API) and X (OAuth1 or OAuth2), once per trigger
• Stores post records and pending/retry state in MongoDB

Repo structure (key files)

• main.py: single-run orchestrator (used by Lambda)
• lambda_handler.py: Lambda entrypoint calling run_once()
• app/
  • config.py: reads env and keywords
  • fetch_reddit.py, fetch_x.py: source fetchers
  • generate.py: OpenAI gpt-4o prompt/generation (one item)
  • post_linkedin.py, post_x.py: posting clients
  • db_mongo.py: MongoDB helpers (pending/posted records)
• serverless.yml: Serverless Framework config (Lambda + EventBridge schedules)

Requirements

• Python 3.9+ locally (Lambda currently set to python3.9 in serverless.yml)
• AWS account + credentials configured for Serverless Framework
• OpenAI API key
• LinkedIn access token with permissions for UGC posts
• X API credentials:
  • Either OAuth1 keys with write permissions (read+write)
  • Or an OAuth2 user access token with tweet.write scope
• MongoDB (Atlas or DocumentDB or local)

Environment variables

Required (match serverless.yml):

• OPENAI_API_KEY
• LINKEDIN_ACCESS_TOKEN
• REDDIT_CLIENT_ID, REDDIT_CLIENT_SECRET
• X_BEARER_TOKEN (for search)
• One of (for posting to X):
  • OAuth1: X_API_KEY, X_API_SECRET, X_ACCESS_TOKEN, X_ACCESS_TOKEN_SECRET
  • OAuth2: X_CLIENT_ID, X_CLIENT_SECRET (to mint tokens externally), and X_OAUTH2_ACCESS_TOKEN for posting
• MongoDB: MONGO_URI, MONGO_DB (default autoposter), MONGO_COLLECTION (default posts)

Optional/unused by core flow (may be present in serverless.yml): LINKEDIN_ID_TOKEN, LINKEDIN_CLIENTID, LINKEDIN_SECRETID, Discord vars.

Local test

Install dependencies and run one cycle locally:

pip install -r requirements.txt
python lambda_handler.py

This prints {"status": "ok"} and executes a full run.

Deploy with Serverless Framework

1. Ensure Serverless is installed and AWS credentials are set.
2. Place your environment variables in a local .env (the config uses useDotenv: true).
3. Install plugins:

sls plugin install -n serverless-python-requirements
sls plugin install -n serverless-dotenv-plugin

4. Deploy:

sls deploy

5. Invoke once to test:

sls invoke -f autoposter

Scheduling

• serverless.yml defines two EventBridge schedules (UTC). Adjust rate: cron(...) or remove the events block if you prefer manual invocation.
• Region defaults to ap-south-1; change provider.region as needed.

serverless.yml explained (concise)

• service/frameworkVersion: project name and Serverless v3.
• useDotenv: true: loads .env into provider.environment for deploys.
• provider:
  • runtime: Python runtime for Lambda (python3.9 set here).
  • region/stage: AWS region and stage; change region to where you deploy.
  • memorySize/timeout: Lambda resources (adjust if needed).
  • environment: all env vars your function reads (OpenAI, LinkedIn, X, Mongo, etc.).
• plugins:
  • serverless-python-requirements: builds Python deps into the bundle (uses Docker if dockerizePip: true).
  • serverless-dotenv-plugin: injects .env into env variables at deploy time.
• custom.pythonRequirements:
  • dockerizePip: set true for native wheels compatibility; false uses local pip.
  • slim/strip: reduce package size by removing extraneous files.
• package.patterns: include everything except caches/dist-info/pyc and old sqlite file.
• functions.autoposter:
  • handler: entry is lambda_handler.handler.
  • events.schedule: two EventBridge cron triggers in UTC; set enabled: true/false or edit the cron to change run times.

Notes and tips

• X rate limits: The fetcher trims keywords and retries with a minimal set. If you still hit limits or see 403, reduce keyword breadth or ensure your app has appropriate access.
• LinkedIn posts with URLs are sent as ARTICLE shares with originalUrl and DataMap-wrapped title.
• MongoDB is used to avoid reposts and to retry pending items automatically on next run.
• Keywords/subreddits are broad by default; override via KEYWORDS env if desired.

Security

• Do not commit secrets. Use .env locally and set environment variables in Lambda/Serverless for production.
• Consider AWS Secrets Manager/SSM for managing sensitive values.