chach

[yaacov]

Summary by CodeRabbit

• New Features

  • Added --resume ID option to resume saved chat sessions; session IDs and copy-pastable resume commands are now printed on exit.
  • Increased conversation history capacity from 80,000 to 160,000 characters for larger context windows.
  • Improved session persistence: full chat history, including tool interactions, is restored on resume.

• Bug Fixes

  • Fixed message duplication when resuming sessions.

• Documentation

  • Updated README with resume session instructions.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR adds session resumption to the MTV-agent TUI, allowing users to save and restore chat sessions via a --resume ID flag. The implementation increases the message history budget, improves history trimming with atomic tool-call grouping, adds server-side message snapshotting for clean persistence, and wires CLI/TUI support for resuming saved chats.

Changes

Session Resume and History Persistence

Layer / File(s)	Summary
Configuration and documentation for new defaults mtv_agent/server/config.py, README.md	Default max_history_chars is doubled from 80,000 to 160,000. README documents the --resume ID option for both mtv-agent run and mtv-tui, and describes session persistence: on exit, session ID and resume command are printed; chat history including tool interactions is cached and fully restored.
History budget and grouping-based trimming mtv_agent/server/agent.py, mtv_agent/server/tools.py	agent.py increases default max_history_chars in both run_stream and _build_messages to 160,000. tools.py adds json import and replaces flat per-message trimming with grouping that treats assistant tool_calls plus following tool messages as atomic, non-splittable units; _truncate default is also increased to 160,000.
Server message snapshotting for persistence mtv_agent/server/agent.py, mtv_agent/server/app.py	agent.py adds _prepare_snapshot helper to remove system prompt, strip metadata keys, and truncate oversized tool content, then emits both _messages_snapshot and content events after final LLM responses. app.py treats snapshot events as authoritative state replacement and avoids duplicating assistant messages when persisting history.
CLI and TUI resume implementation mtv_agent/cli.py, mtv_agent/tui/app.py	cli.py adds --resume ID argument and passes it to MTVApp; after the app exits, prints resume instructions using the session ID prefix. tui/app.py accepts optional resume_id, calls _resume_chat on mount, filters assistant-only messages during replay (skipping tool-call turns), and the main CLI entry point integrates the --resume argument and session recovery flow.

🎯 3 (Moderate) | ⏱️ ~25 minutes

🐰 With whiskers twitching bright,
Sessions saved, resuming right,
History grouped in atomic pairs,
Snapshots clean—the rabbit declares!
Chat flows on from where it starts,
No lost context in these parts! 🎉

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 55.56% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Title check	❓ Inconclusive	The title 'chach' is vague and non-descriptive, providing no meaningful information about the changeset's purpose or scope.	Use a clear, descriptive title that summarizes the main change, such as 'Add session resume support with chat history persistence' or 'Implement --resume CLI option for saved TUI sessions'.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch cache

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@mtv_agent/server/agent.py`:
- Around line 173-177: The dict comprehension that builds cleaned incorrectly
uses "or" and should use "and": locate the comprehension that assigns cleaned
from msg (symbols: cleaned, msg, _SNAPSHOT_STRIP_KEYS) and change the condition
from "if k not in _SNAPSHOT_STRIP_KEYS or v is not None" to "if k not in
_SNAPSHOT_STRIP_KEYS and v is not None" so keys listed in _SNAPSHOT_STRIP_KEYS
are always removed while keeping only non-None values for other keys.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 09509645-5e1b-4c37-859b-4a20f73876a9

📥 Commits

Reviewing files that changed from the base of the PR and between 6820526 and cce6381.

⛔ Files ignored due to path filters (1)

• uv.lock is excluded by !**/*.lock

📒 Files selected for processing (7)

• README.md
• mtv_agent/cli.py
• mtv_agent/server/agent.py
• mtv_agent/server/app.py
• mtv_agent/server/config.py
• mtv_agent/server/tools.py
• mtv_agent/tui/app.py

[coderabbitai]

⚠️ Potential issue | 🔴 Critical | ⚡ Quick win

Critical logic error: or should be and in the dict comprehension.

The condition if k not in _SNAPSHOT_STRIP_KEYS or v is not None incorrectly keeps keys that should be stripped. For example, if "annotations" has a non-None value, the condition evaluates to False or True = True, so the key is kept—but annotations is in _SNAPSHOT_STRIP_KEYS and should always be removed. Similarly, keys not in strip_keys but with None values are incorrectly kept.

🐛 Proposed fix

         cleaned = {
             k: v
             for k, v in msg.items()
-            if k not in _SNAPSHOT_STRIP_KEYS or v is not None
+            if k not in _SNAPSHOT_STRIP_KEYS and v is not None
         }

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	cleaned = {
	k: v
	for k, v in msg.items()
	if k not in _SNAPSHOT_STRIP_KEYS or v is not None
	}
	cleaned = {
	k: v
	for k, v in msg.items()
	if k not in _SNAPSHOT_STRIP_KEYS and v is not None
	}

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@mtv_agent/server/agent.py` around lines 173 - 177, The dict comprehension
that builds cleaned incorrectly uses "or" and should use "and": locate the
comprehension that assigns cleaned from msg (symbols: cleaned, msg,
_SNAPSHOT_STRIP_KEYS) and change the condition from "if k not in
_SNAPSHOT_STRIP_KEYS or v is not None" to "if k not in _SNAPSHOT_STRIP_KEYS and
v is not None" so keys listed in _SNAPSHOT_STRIP_KEYS are always removed while
keeping only non-None values for other keys.