[FEATURE] Add --pipe flag for Unix pipeline integration

[jwesleye]

---

name: Feature Request
about: Suggest an idea for this project
title: '[FEATURE] Add --pipe flag for Unix pipeline integration'
labels: enhancement
assignees: ''

Feature Description

Add a --pipe flag that enables true stdin/stdout mode for seamless integration with Unix pipelines and automation scripts.

Problem/Motivation

Is your feature request related to a problem?
The current interactive mode is great for human use but doesn't work well with shell scripts, CI/CD pipelines, or Unix tool chains. I need a way to use chat_loop as a filter in standard Unix pipelines.

Example: "I'm frustrated when I need to process multiple queries in a script because the interactive prompts and formatting break pipeline composition."

Proposed Solution

Add a pipe mode that reads queries from stdin (one per line) and writes responses to stdout with minimal formatting.

Example:

# Basic pipeline
echo "What is 2+2?" | chat_loop myagent --pipe
# Output: 4

# Process multiple queries
cat queries.txt | chat_loop myagent --pipe > responses.txt

# Chain with other tools
curl https://api.example.com/data |
  jq -r '.questions[]' |
  chat_loop myagent --pipe |
  tee results.txt

# Parallel processing
parallel -j4 "chat_loop myagent --pipe" < queries.txt > all_responses.txt

# Integration with data processing
awk '{print "Summarize: " $0}' logfile.txt |
  chat_loop myagent --pipe > summaries.txt

Alternatives Considered

1. Expect scripts - Too fragile, breaks with format changes
2. API wrapper - Requires additional service, overkill for simple scripts
3. Separate tool - Would fragment the ecosystem
4. --quiet flag - Not enough, need proper stdin handling

Benefits

Who would benefit from this feature and how?

• DevOps engineers: Integrate agents into CI/CD pipelines
• Data scientists: Batch processing of datasets
• Automation scripts: Reliable, scriptable agent interaction
• Unix enthusiasts: Follows Unix philosophy (do one thing well, composable)

Implementation Ideas

Behavior in pipe mode:

• Disable all interactive prompts and readline
• Read one query per line from stdin
• Write one response per line to stdout
• Suppress banners, status bars, colors (unless --pipe-color)
• Exit after processing all input (no interactive loop)
• Write errors to stderr only
• Exit code 0 on success, non-zero on failure

Optional enhancements:

# Keep delimiters for multi-line responses
chat_loop myagent --pipe --delimiter "---"

# JSON mode for structured output
chat_loop myagent --pipe --json
# Output: {"query": "...", "response": "...", "tokens": 123, "cost": 0.01}

# Batch processing with status
chat_loop myagent --pipe --progress
# Writes progress to stderr: [1/100] Processing query...

Suggested implementation approach:

def pipe_mode(agent, config):
    """Process stdin queries in batch mode."""
    try:
        for line in sys.stdin:
            query = line.strip()
            if not query:
                continue

            try:
                response = agent(query)
                # Extract just the text response
                text = extract_response_text(response)
                print(text, flush=True)
            except Exception as e:
                print(f"ERROR: {e}", file=sys.stderr)
                sys.exit(1)
    except KeyboardInterrupt:
        sys.exit(130)  # Standard Unix SIGINT exit code

Additional Context

This follows the Unix philosophy of composable tools:

• Each program does one thing well
• Programs work together via stdin/stdout
• Text streams as universal interface

Similar to:

• jq for JSON processing
• sed/awk for text transformation
• curl with -s flag
• ffmpeg with -i - for stdin

Would make chat_loop a first-class citizen in shell tooling.

Example use cases:

# Log analysis
tail -f app.log | grep ERROR | chat_loop analyst --pipe |
  notify "New error analysis"

# Code review automation
git diff | chat_loop reviewer --pipe > review_comments.txt

# Content generation
cat topics.txt | chat_loop writer --pipe > articles/

# Data enrichment
psql -t -c "SELECT description FROM products" |
  chat_loop classifier --pipe |
  psql -c "UPDATE products SET category = stdin"

Priority

How important is this feature to you?

• Critical - Blocking my workflow
• High - Would significantly improve my experience
• Medium - Nice to have
• Low - Minor improvement

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity.
It will be closed in 14 days if no further activity occurs.

If this issue is still relevant, please:

• Comment to keep it open
• Remove the "status: stale" label
• Provide additional information or updates

Thank you for your contributions!

[github-actions]

This issue has been automatically closed due to inactivity.

If you believe this issue is still relevant:

• Feel free to reopen it with additional context
• Create a new issue referencing this one

Thank you for understanding!