mayros batch

Process multiple prompts in parallel with configurable concurrency. Input can be a JSONL file, a text file with --- separators, or piped from stdin. Results are streamed as JSON-lines to stdout or written to a file.

Subcommands

batch run [file]

Process a file of prompts in parallel. Each prompt is sent to the Gateway as an independent chat request.

Arguments:

ArgumentDescription
[file]Input file (JSONL or text with --- separators). Use - for stdin.

Options:

FlagDescriptionDefault
-c, --concurrency <n>Max concurrent prompts (1-16)4
-o, --output <file>Write results to file instead of stdout—
--jsonOutput results as JSON-linesfalse
--session <key>Session key prefix (each item gets a unique suffix)—
--thinking <level>Thinking level for all prompts—
--timeout <ms>Per-prompt timeout in milliseconds120000
--url <url>Gateway WebSocket URL—
--token <token>Gateway auth token—
--password <password>Gateway password—

Input formats:

  • JSONL — one JSON object per line with at least a prompt field. Optional id and context fields.
  • Text — prompts separated by --- on its own line. IDs are auto-assigned (1, 2, 3...).
  • stdin — pipe data to mayros batch run -.

Format is auto-detected: if the first non-empty line starts with {, the file is parsed as JSONL; otherwise as ----separated text.

Output format:

Each result is a JSON object with the following fields:

json
{
  "id": "1",
  "status": "ok",
  "response": "The assistant response...",
  "durationMs": 1234
}

On errors, status is "error" and the error field contains the reason.

batch status

Show batch processing capabilities and usage examples.

Examples

bash
# Process a JSONL file with 8 concurrent prompts
mayros batch run prompts.jsonl --concurrency 8

# Process a text file and write results to a file
mayros batch run prompts.txt --output results.jsonl

# Pipe a single prompt from stdin
echo '{"prompt":"Explain monads"}' | mayros batch run -

# Set a custom timeout and thinking level
mayros batch run prompts.jsonl --timeout 60000 --thinking high

# Use a specific gateway
mayros batch run prompts.jsonl --url ws://localhost:8080 --token mytoken

JSONL input example (prompts.jsonl):

json
{"id": "q1", "prompt": "What is dependency injection?"}
{"id": "q2", "prompt": "Explain the observer pattern", "context": "In TypeScript"}
{"id": "q3", "prompt": "Compare REST and GraphQL"}

Text input example (prompts.txt):

text
What is dependency injection?
---
Explain the observer pattern
---
Compare REST and GraphQL
  • Headless mode — single-prompt non-interactive mode
  • code — interactive coding session