feat(cli): introduce AI-agent friendly CLI with JSON output, non-interactive mode, and standardized exit behavior

🎯 Objective

This MR enhances the CLI to support AI agents, automation workflows, and headless execution by introducing structured outputs, eliminating interactive dependencies, and standardizing behavior.

---

✨ Key Features

1. 🧾 JSON Output Support

• Added global --json flag
• Outputs machine-readable structured JSON
• Ensures:
  • No rich tables / ANSI formatting
  • No progress bars (tqdm)
  • Clean output for parsing by scripts/LLMs

---

2. 🤖 Non-Interactive Mode

• Added global --non-interactive flag
• Disables all prompts (input(), confirmations, selections)
• Behavior:
  • Uses defaults when safe
  • Fails gracefully with clear errors when required input is missing

---

3. 📊 Structured Job Output

• Standardized JSON schema for job data:

{
"index": 1,
"mode": "upload",
"status": "completed",
"progress": {
"completed": 1,
"failed": 0,
"total": 1
},
"last_updated": "timestamp",
"completed_files": [],
"failed_files": []
}

---

4. 🔁 Backward Compatibility

• Existing CLI behavior remains unchanged by default
• Rich UI (tables, colors) still works without --json

---

5. ⚠️ Exit Code Standardization

Scenario	Exit Code
Success	0
Validation Error (e.g., invalid job index)	1
API / Network / System Error	2

---

6. 📤 Output Stream Separation

• stdout → JSON output only
• stderr → logs, warnings, errors
• Prevents JSON corruption in automation pipelines

---

🛠️ Implementation Highlights

• Added global flags via Typer context
• Introduced centralized output handling (output.py)
• Refactored job commands (list, show) to support dual output modes
• Added guard for interactive prompts in non-interactive mode
• Updated error handling to align with exit codes
• Ensured consistent behavior across commands

---

✅ Testing & Verification

✔ Manual Tests

• corpus-client job list --json
• corpus-client job show 1 --json
• corpus-client job show 999 --json → returns JSON error + exit code 1
• corpus-client job --non-interactive

✔ Output Validation

• JSON is clean and parseable (no extra logs in stdout)
• stderr correctly captures warnings/errors

✔ Exit Codes

• Verified using echo $?

---

📌 Impact

• Enables AI/LLM integration
• Supports automation pipelines & scripting
• Makes CLI production-ready for headless environments

---

🔮 Future Enhancements (Optional)

• Add --format yaml
• Extend JSON output to all commands
• Add schema validation for outputs