askOllama(1)

NAME

askOllama - Command‐line interface for batch querying Ollama models with RAG support

SYNOPSIS

   askOllama [-o outfile] [-p personafile] [--rag RAGDIR] prompt [arg]

DESCRIPTION

askOllama is a command‐line tool for sending prompts and arguments to Ollama models and retrieving responses. It supports batch queries, persona files, Retrieval‐Augmented Generation (RAG) directories, and multiple output formats.

askOllama supports Retrieval-Augmented Generation (RAG) by indexing documents in a specified directory or a single file. The vector database used for RAG is isolated in a hidden subdirectory `.rag_db` inside the RAG path (or the file's parent directory), and that directory is excluded from the corpus so it does not interfere with which files are indexed.

Prompts and arguments can be read from files: if the positional prompt or arg looks like a file path, it will be parsed. Multiple prompts and multiple arguments are combined in a Cartesian product (N×M) and executed, with results written in order.

OPTIONS

   -o outfile
          Specify an output file for results. Supported  formats:  .jsonl,
          .csv,  .rtf,  or  plain text. If omitted, results are printed to
          stdout.

   -p personafile
          Specify a file containing a persona/system prompt to set for the
          model.

   --rag RAGDIR
          Directory (or a single file) to build a RAG index from. The index is stored in `.rag_db` under the
          RAG directory. The `.rag_db` directory is ignored while crawling.

   prompt A prompt string or a file containing prompts (one per line).

   arg    An  optional  argument  string or file containing arguments (one
          per line).

   ‐h     Show usage/options summary.

   ‐‐help  Show this help/man page.

   --verbose  Enable verbose logging (DEBUG level). Default logging is INFO.

SUPPORTED FILETYPES

   RAG context files:
          PDF (.pdf), text (.txt, .log, .py, .json), HTML  (.html,  .htm),
          Markdown  (.md, .markdown), RTF (.rtf), Word (.doc, .docx), Pow‐
          erPoint (.ppt, .pptx), Excel (.xls, .xlsx)

   Prompt files:
          Text (.txt), JSONL (.jsonl), CSV (.csv), RTF (.rtf)

   Argfiles:
          Text (.txt), JSONL (.jsonl), CSV (.csv), RTF (.rtf)

   Outfiles:
          JSONL (.jsonl), CSV (.csv), RTF (.rtf), plain text (.txt)

EXAMPLES

Query with a single prompt and print the result to stdout:

   > askOllama "Who is the 47th president of the USA?"

Query with prompts from a file and save results to JSONL:

   > askOllama -o outfile.jsonl prompts.txt

Query with persona and RAG context:

   > askOllama -o outfile.txt -p Grogu.txt --rag RAGDIR prompts.txt argfile.txt

INSTALLATION

askOllama is an all-in-one Python executable. Just put it in your executable path or current working directory. However, like most Python programs, askOllama requires dependencies to be loaded via pip.

python -m pip install --upgrade pip setuptools wheel

# Core
pip install \
  requests \
  langchain langchain-community langchain-chroma langchain-ollama chromadb \
  unstructured pypdf \
  python-docx python-pptx openpyxl

ENVIRONMENT

Heads-up: you still need an Ollama server running locally (default http://localhost:11434), and --llm overrides OLLAMA_MODEL if you set that env var.

   OLLAMA_API_BASE
          Base URL for the Ollama API (default: http://localhost:11434).

   OLLAMA_MODEL
          Model name for Ollama (default: llama2).

AUTHOR

Binary Systems, Inc.
http://www.binary-systems.com