pstexplorer

A CLI tool to explore and extract data from Outlook PST files.

Installation

Check the instructions on the latest release page.

Usage

pstexplorer <COMMAND>

Commands:
  list    List all emails in a PST file
  search  Search emails in a PST file by query string (matches from, to, cc, body)
  browse  Browse PST file contents in a TUI
  stats   Print statistics about a PST file
  export  Export a PST file to a SQLite database
  llm     LLM-powered commands (embed emails, ask questions)

Features

stats

Print a summary of the PST file: folder count, email/calendar/contact/task/note counts, attachment count, and date range.

PST Statistics: "testdata/testPST.pst"
  Folders:          2
  Total items:      6
  Emails:           6
  Attachments:      0
  Earliest message: 2014-02-24 21:14:34 UTC
  Latest message:   2014-02-26 12:20:19 UTC

list

List all emails with subject, sender, recipient, and date. Supports --format csv|tsv|json for structured output and --limit to cap the number of entries.

search

Case-insensitive full-text search across from, to, cc, and body fields. Supports the same --format options as list.

$ pstexplorer search --format json -- testdata/testPST.pst "tika"
[
  {
    "folder": "Début du fichier de données Outlook",
    "subject": "\u0001\u0001[jira] [Resolved] (TIKA-1249) Vcard files detection",
    "from": "Nick Burch (JIRA)",
    "to": "dev@tika.apache.org",
    "cc": "",
    "date": "2014-02-26 12:20:25 UTC"
  },
  {
    "folder": "Début du fichier de données Outlook",
    "subject": "\u0001\u0001[jira] [Commented] (TIKA-1250) Process loops infintely processing a CHM file",
    "from": "Gary Murphy (JIRA)",
    "to": "dev@tika.apache.org",
    "cc": "",
    "date": "2014-02-26 12:12:25 UTC"
  }
]

export

Export folders and messages to a SQLite database for further analysis. Use --output to set the database path and --limit to cap the number of exported messages.

Tip

export to a SQLite db and then use uvx datasette to visually browse the data

browse

Interactive terminal UI for navigating folders and reading messages.

llm embed

Index emails into a ChromaDB vector database. Embeddings are generated via any OpenAI-compatible API — locally with Ollama or remotely with OpenAI.

The collection name defaults to the PST filename stem (e.g. testPST.pst → testPST).

# with Ollama
pstexplorer llm embed testPST.pst \
  --embedding-url http://localhost:11434/v1 \
  --embedding-model nomic-embed-text

# with OpenAI
pstexplorer llm embed testPST.pst \
  --embedding-url https://api.openai.com/v1 \
  --embedding-key sk-... \
  --embedding-model text-embedding-3-small

llm ask

Ask a natural language question about your emails. Relevant messages are retrieved from ChromaDB and passed as context to a chat model.

Important

The --embedding-model must match the model used during llm embed.

# with Ollama
pstexplorer llm ask --collection testPST \
  --embedding-url http://localhost:11434/v1 \
  --embedding-model nomic-embed-text \
  --llm-url http://localhost:11434/v1 \
  --llm-model llama3.2 \
  "who sent me invoices last year?"

# with OpenAI
pstexplorer llm ask --collection testPST \
  --embedding-url https://api.openai.com/v1 \
  --embedding-model text-embedding-3-small \
  --llm-url https://api.openai.com/v1 \
  --llm-model gpt-4o-mini \
  "summarise the thread about the budget"

API keys can also be set via environment variables to keep them out of shell history:

export EMBEDDING_API_KEY=sk-...
export LLM_API_KEY=sk-...

Use --n-results (default: 5) to control how many emails are retrieved as context.

Alternative: llmquery/query.py

llmquery/query.py is a standalone script that does the same RAG query using the Python chromadb and ollama libraries directly. It requires a locally running Ollama instance and is self-contained with inline dependency metadata — only uv is needed.

./llmquery/query.py "who sent me invoices in 2023?"
./llmquery/query.py --collection testPST "what is the date and time of the oldest email?"
./llmquery/query.py --collection testPST --n-results 10 "which emails contain source code?"

Datasette plugin

The plugins/ folder includes a Datasette plugin which makes browsing emails a bit easier, by providing detail views with next/prev buttons and a layout which separates header data.

After exporting an SQLite db with pstexplorer export my.pst run:

uvx datasette serve my.db --plugins-dir plugins/

Run the playwright-driven integration test for this datasette plugin with:

uv run --with datasette --with pytest --with pytest-playwright pytest