reheat 0.0.4

Python CLI for SEO intent analysis and content gap discovery.

Reheat

Semantic Intent Clustering + Content Gap Analysis for Google Search Console

Reheat turns the flat, overwhelming list of queries in Google Search Console into clear, actionable insights about what your audience actually wants.

It pulls your GSC data, enriches it with related searches and People Also Ask, clusters queries by semantic intent, and shows you high-value topics and content opportunities in a local web dashboard.

Screenshots from my own site (bayis.co.uk)

Why people use Reheat

• Discover which topics your content already serves across many queries
• Find high-potential content gaps ranked by opportunity
• Get a practical content expansion schedule
• Understand real user intent instead of guessing
• Everything runs locally (or with Postgres)

Built and used daily by me on my own technical/ML content site.

---

Quick Start

pip install reheat

Full setup instructions (including Google Search Console OAuth and SerpAPI key) -> GETTING_STARTED.md

Once set up, the basic flow is:

reheat fetch && reheat enrich && reheat analyse && reheat serve

Then open localhost:8000

Key Features

• Google Search Console data import
• SerpAPI enrichment (related searches + PAA)
• Local embeddings + semantic clustering
• LLM cluster labelling (OpenAI / Anthropic / Marigold)
• High-value topic detection
• Content expansion recommendations
• Interactive intent map (UMAP scatter plot)
• Local FastAPI web dashboard

Continue reading for full setup, CLI reference, and architecture

---

Getting started

1. Install

pip install reheat
# or in a virtualenv:
python -m venv venv && source venv/bin/activate
pip install reheat

See GETTING_STARTED.md for a full walkthrough including Google Cloud setup and SerpAPI configuration.

2. Start a postgres instance

reheat uses postgres to store runs, enrichments, and report data.

docker run -d \
  --name reheat-pg \
  --rm \
  -e POSTGRES_USER=reheat \
  -e POSTGRES_PASSWORD=reheat \
  -e POSTGRES_DB=reheat \
  -p 5432:5432 \
  postgres:16

3. Set environment variables

# database
export DATABASE_URL="postgresql://reheat:reheat@localhost:5432/reheat"

# google search console (OAuth2 Desktop app credentials)
export GOOGLE_CLIENT_SECRETS_PATH="/path/to/client_secrets.json"
export GOOGLE_TOKEN_PATH="/path/to/token.json"

# serpapi (optional, for related search enrichment)
export SERPAPI_KEY="your-serpapi-key"

# llm provider (one of the following)
export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."

The Google credentials file must be an OAuth 2.0 Client ID of type Desktop app. In Google Cloud Console go to APIs and Services > Credentials > Create Credentials > OAuth 2.0 Client ID, select Desktop app, and download the JSON file. Service account keys and web application credentials will not work.

GOOGLE_TOKEN_PATH is where reheat writes the OAuth token after the first consent flow. Point it to a persistent location. The browser consent flow runs automatically on the first reheat fetch and is not required again until the token expires.

4. Register sources

# google search console
reheat sources create \
  --source-type google_search_console \
  --domain yourdomain.com \
  --days 180

# serpapi (optional)
reheat sources create \
  --source-type serp \
  --domain google

The --days flag sets the GSC lookback window (default 90, maximum ~480). The --domain flag on the serp source sets the search engine. Supported values: google, youtube, google_patents, google_news.

5. Run the pipeline

reheat fetch
reheat enrich
reheat analyse
reheat serve

Open http://localhost:8000.

The four commands cover the full pipeline. Individual steps are also available if you need to re-run a specific stage:

reheat fetch                      # pull queries from Google Search Console
reheat enrich adjacent            # fetch related searches via SerpAPI
reheat enrich tags                # auto-tag queries
reheat enrich embed               # generate embeddings
reheat enrich cluster             # cluster by semantic intent
reheat analyse summarise          # label clusters with an LLM
reheat analyse opportunities      # score content gaps
reheat analyse schedule           # generate content schedule
reheat analyse overview           # generate narrative summary
reheat project create             # compute UMAP projection
reheat report scatter create      # build scatter plot data
reheat report summary create      # build summary panel data
reheat report coverage create     # build coverage table data
reheat serve                      # start the web interface

---

Inference providers

reheat analyse labels intent clusters and generates a content schedule using an LLM. Set one of the following environment variables.

export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."

Marigold is a private inference API built by Bay Information Systems. Configure it in reheat user settings:

reheat config set --key marigold_endpoint --value https://api.marigold.run
reheat config set --key marigold_api_key --value <key>

---

CLI reference

reheat fetch
reheat enrich [adjacent | tags | embed | cluster]
reheat analyse [summarise | opportunities | schedule | overview]
reheat project [create | read]
reheat report [scatter | summary | coverage | opportunities | overlaps] [create | read]
reheat serve

reheat sources [create | list | show | update | delete]
reheat runs [list | show | delete]
reheat config [show | set]
reheat status

Pass --json before any command for machine-readable output:

reheat --json sources list
reheat --json runs list

---

Architecture

reheat has three layers.

Commands in reheat/commands/ are the single source of truth for the application surface. Each command is a Python function decorated with @command, registered in a central registry, and exposed automatically through both the CLI and the HTTP API.

Pipeline functions in reheat/pipeline/ are pure data transforms: embedding, clustering, gap analysis, report building. No persistence, no side effects.

Persistence uses dynawrap, a lightweight key-value library with identical interfaces over PostgreSQL and DynamoDB. Tables are passed at call time; models are backend-agnostic. The backend is selected from DATABASE_URL at startup.

The web interface is a static SPA served by FastAPI. All pages share a single stylesheet and a common api.js module that is the single source of truth for API endpoint calls.

---

Optional dependencies

pip install reheat[openai]       # OpenAI LLM support
pip install reheat[anthropic]    # Anthropic LLM support
pip install reheat[postgres]     # PostgreSQL backend (psycopg2)
pip install reheat[all]          # all of the above

---

License

MIT. See LICENSE.

Built by Edward Grundy at Bay Information Systems.