Manual compaction counting fix + cli cleanup #5480

katzdave · 2025-10-30T16:57:56Z

Token counts updated in the stream immediately after compaction.

Remove lots of unused code from compaction; only return the provider usage.

Modify update_session_metrics to support using summarization usage to estimate new context window (with system prompt token estimate).

Remove lots of custom code in CLI around token counting + manual summarization trigger

Rename /summarize -> /compact.

Copilot

Pull Request Overview

This PR refactors the conversation compaction feature by simplifying the return type and centralizing token tracking logic. The changes rename "summarize" to "compact" throughout the codebase for consistency, remove redundant token count tracking in the compaction flow, and integrate metrics updates directly into the agent response handling.

Key changes:

Simplified compact_messages return type from (Conversation, Vec<usize>, Option<ProviderUsage>) to (Conversation, ProviderUsage)
Centralized token tracking logic in update_session_metrics with special handling for compaction usage
Renamed "/summarize" command to "/compact" with backward compatibility for the deprecated command

Reviewed Changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
crates/goose/src/context_mgmt/mod.rs	Simplified compact_messages return type and removed token count tracking logic
crates/goose/src/agents/reply_parts.rs	Added compaction-aware token tracking with SYSTEM_PROMPT_TOKEN_ESTIMATE and is_compaction_usage parameter
crates/goose/src/agents/agent.rs	Updated compact_messages call sites to use new signature and added session metrics updates
crates/goose/src/agents/mod.rs	Exported MANUAL_COMPACT_TRIGGER constant for CLI usage
crates/goose-cli/src/session/mod.rs	Refactored manual compaction to use agent response processing instead of direct API calls
crates/goose-cli/src/session/input.rs	Renamed InputResult::Summarize to InputResult::Compact with deprecation warning for /summarize

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-11-03T17:35:12Z

crates/goose/src/agents/reply_parts.rs

 use crate::session::SessionManager;
 use rmcp::model::Tool;

+const SYSTEM_PROMPT_TOKEN_ESTIMATE: i32 = 5000;


This magic number lacks documentation explaining why 5000 tokens is used as an estimate for the system prompt. Add a comment explaining the rationale behind this value and consider making it configurable if system prompt size varies significantly across different use cases.

Suggested change

const SYSTEM_PROMPT_TOKEN_ESTIMATE: i32 = 5000;

/// Estimated token count for the system prompt.

/// The default value of 5000 tokens is chosen based on typical prompt sizes and model limitations.

/// If your use case requires a different estimate, you can override this value by setting the

/// `SYSTEM_PROMPT_TOKEN_ESTIMATE` environment variable.

fn system_prompt_token_estimate() -> i32 {

std::env::var("SYSTEM_PROMPT_TOKEN_ESTIMATE")

.ok()

.and_then(|v| v.parse::<i32>().ok())

.unwrap_or(5000)

}

jamadeo · 2025-11-03T18:12:53Z

crates/goose/src/agents/reply_parts.rs

            accumulate(session.accumulated_output_tokens, usage.usage.output_tokens);

+        let (current_total, current_input, current_output) = if is_compaction_usage {
+            // After compaction: summary output becomes new input context


after manual compaction, should this not really just go to zero? have we sent the manually compacted conversation to the model yet or only to the client? although I could see how that would be confusing to a user

it would be good to avoid a hard coded system prompt estimate, just because that could really be all over the place

I suppose we could do something like use the token counts from the first message you send, but that would be complicated to do

Yeah I think this one is more of a question of what we want to show the user. Let's try dropping it to the summary size, there will be a bit of whiplash for the user when the system prompt re-enters on the next message, but I don't love that constant either.

jamadeo · 2025-11-03T18:17:44Z

crates/goose/src/context_mgmt/mod.rs

 /// * A tuple containing:
 ///   - `Conversation`: The compacted messages
-///   - `Vec<usize>`: Token counts for each message
-///   - `Option<ProviderUsage>`: Provider usage from summarization


nice, I saw we didn't use this. There's some code on the frontend you could clean up with this too

ack will take another pass on the frontend too.

…se into dkatz/manual-compact-fix-usage * 'dkatz/manual-compact-fix-usage' of github.com:block/goose: (35 commits) Fix image processing (#5544) docs: AI attribution for PRs (#5547) chore(tests/mcp): testing for MCP sampling (#5456) docs: adding HOWTOAI.md (#5533) added configuration content, also added signoff, fix merging issue with another commit by creating a clean branch. removed and closed commits that caused signoff issues. (#5519) Fixes Gemini API parse issue by converting nullable type arrays to single types in tool schemas (#5530) Troubleshooting diagnostics doc (#5526) fix link to Ollama FAQ (#5531) docs: remove speech-mcp (#5514) fix: adds ProviderRetry to openai provider (#5518) docs: extensions directory minor updates (#5466) Docs/json recipe support (#5492) docs: recipe buttons (#5507) Improve system theme detection and fallback (#5427) [Autovisualiser] remove unnecessary content from mermaid HTML template (#5505) Improve subagents docs (#5484) FIX: prefer linux in WSL and add INSTALL_OS override for CLI (#5215) Propagate session ID in LLM and MCP requests (#5165) feat: YT Short for Canva MCP + goose (#5495) Change Recipes Test Script (#5457) ...

Copilot

Pull Request Overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-11-03T20:52:04Z

crates/goose/src/agents/reply_parts.rs


+        let (current_total, current_input, current_output) = if is_compaction_usage {
+            // After compaction: summary output becomes new input context
+            let new_input = usage.usage.output_tokens.map(|out| out);


The identity closure map(|out| out) is redundant. Consider simplifying to just usage.usage.output_tokens directly.

Suggested change

let new_input = usage.usage.output_tokens.map(|out| out);

let new_input = usage.usage.output_tokens;

Copilot · 2025-11-03T20:52:05Z

crates/goose-cli/src/session/input.rs

-        s if s == CMD_SUMMARIZE => Some(InputResult::Summarize),
+        s if s == CMD_COMPACT => Some(InputResult::Compact),
+        s if s == CMD_SUMMARIZE_DEPRECATED => {
+            println!("{}", console::style("⚠️  Note: /summarize has been renamed to /compact and will be removed in a future release.").yellow());


[nitpick] The deprecation message string is very long (over 100 characters). Consider breaking it into multiple lines or storing it as a constant for better readability and maintainability.

* 'main' of github.com:block/goose: (21 commits) Manual compaction counting fix + cli cleanup (#5480) chore(deps): bump prismjs and react-syntax-highlighter in /ui/desktop (#5549) fix: remove qwen3-coder from provider/mcp smoke tests (#5551) fix: do not build unsigned desktop app bundles on every PR in ci. add manual option. (#5550) fix: update Husky prepare script to v9 format (#5522) Fix 404 for responsible coding guide (#5543) fix hermit `text file busy` issues on linux (#5372) Fix image processing (#5544) docs: AI attribution for PRs (#5547) chore(tests/mcp): testing for MCP sampling (#5456) docs: adding HOWTOAI.md (#5533) added configuration content, also added signoff, fix merging issue with another commit by creating a clean branch. removed and closed commits that caused signoff issues. (#5519) Fixes Gemini API parse issue by converting nullable type arrays to single types in tool schemas (#5530) Troubleshooting diagnostics doc (#5526) fix link to Ollama FAQ (#5531) docs: remove speech-mcp (#5514) fix: adds ProviderRetry to openai provider (#5518) docs: extensions directory minor updates (#5466) Docs/json recipe support (#5492) docs: recipe buttons (#5507) ...

* main: (85 commits) improve linux tray icon support (#5425) feat: log rotation (#5561) use app.isPackaged instead of checking for node env development (#5465) disable RPM build-ID generation to prevent package conflicts (#5563) Add Diagnostics Info to Q&A and Bug Report Templates (#5565) fix: improve server error messages to include HTTP status code (#5532) improvement: add useful error message when attempting to use unauthenticated cursor-agent (#5300) fix: unblock acp via databricks (#5562) feat: add --output-format json flag to goose run command (#5525) Sessions required (#5548) feat: add grouped extension loading notification (#5529) we should run this on main and also test open models at least via ope… (#5556) info: print location of sessions.db via goose info (#5557) chore: remove yarn usage from documentation (#5555) cli: adjust default theme to address #1905 (#5552) Manual compaction counting fix + cli cleanup (#5480) chore(deps): bump prismjs and react-syntax-highlighter in /ui/desktop (#5549) fix: remove qwen3-coder from provider/mcp smoke tests (#5551) fix: do not build unsigned desktop app bundles on every PR in ci. add manual option. (#5550) fix: update Husky prepare script to v9 format (#5522) ...

* main: (54 commits) add clippy warning for string_slice (#5422) improve linux tray icon support (#5425) feat: log rotation (#5561) use app.isPackaged instead of checking for node env development (#5465) disable RPM build-ID generation to prevent package conflicts (#5563) Add Diagnostics Info to Q&A and Bug Report Templates (#5565) fix: improve server error messages to include HTTP status code (#5532) improvement: add useful error message when attempting to use unauthenticated cursor-agent (#5300) fix: unblock acp via databricks (#5562) feat: add --output-format json flag to goose run command (#5525) Sessions required (#5548) feat: add grouped extension loading notification (#5529) we should run this on main and also test open models at least via ope… (#5556) info: print location of sessions.db via goose info (#5557) chore: remove yarn usage from documentation (#5555) cli: adjust default theme to address #1905 (#5552) Manual compaction counting fix + cli cleanup (#5480) chore(deps): bump prismjs and react-syntax-highlighter in /ui/desktop (#5549) fix: remove qwen3-coder from provider/mcp smoke tests (#5551) fix: do not build unsigned desktop app bundles on every PR in ci. add manual option. (#5550) ...

Signed-off-by: fbalicchia <[email protected]>

Signed-off-by: Blair Allan <[email protected]>

katzdave added 5 commits October 30, 2025 11:10

broadcast summarization usage

c0e2825

update usage to not be optional

214ac1a

Cleanup manual compact cli

1488502

summarize -> compact

29912ad

compaction session metrics fix

6c593dc

katzdave changed the title ~~Compaction counting fix + cli cleanup~~ Manual compaction counting fix + cli cleanup Oct 30, 2025

katzdave marked this pull request as ready for review October 30, 2025 16:58

katzdave added 2 commits October 30, 2025 12:59

useless test

b323e98

rm temporal binary?

a0f1265

katzdave requested a review from jamadeo October 30, 2025 19:03

Merge branch 'main' into dkatz/manual-compact-fix-usage

23ca07c

Copilot AI review requested due to automatic review settings November 3, 2025 17:33

Copilot AI reviewed Nov 3, 2025

View reviewed changes

jamadeo approved these changes Nov 3, 2025

View reviewed changes

katzdave added 3 commits November 3, 2025 15:33

rm system prompt estimate

5649ab8

fmt

bf6ea3d

Copilot AI review requested due to automatic review settings November 3, 2025 20:49

Copilot AI reviewed Nov 3, 2025

View reviewed changes

clippy fix

1deff33

katzdave merged commit 9466d92 into main Nov 3, 2025
14 checks passed

github-actions bot mentioned this pull request Nov 5, 2025

chore(release): release version 1.13.0 (minor) #5582

Merged

fbalicchia pushed a commit to fbalicchia/goose that referenced this pull request Nov 7, 2025

Manual compaction counting fix + cli cleanup (block#5480)

3cf9a95

Signed-off-by: fbalicchia <[email protected]>

BlairAllan pushed a commit to BlairAllan/goose that referenced this pull request Nov 29, 2025

Manual compaction counting fix + cli cleanup (block#5480)

f7072db

Signed-off-by: Blair Allan <[email protected]>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Manual compaction counting fix + cli cleanup #5480

Manual compaction counting fix + cli cleanup #5480

Uh oh!

katzdave commented Oct 30, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Nov 3, 2025

Uh oh!

jamadeo Nov 3, 2025

Uh oh!

katzdave Nov 3, 2025

Uh oh!

jamadeo Nov 3, 2025

Uh oh!

katzdave Nov 3, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Nov 3, 2025

Uh oh!

Copilot AI Nov 3, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

-const SYSTEM_PROMPT_TOKEN_ESTIMATE: i32 = 5000;
+/// Estimated token count for the system prompt.
+/// The default value of 5000 tokens is chosen based on typical prompt sizes and model limitations.
+/// If your use case requires a different estimate, you can override this value by setting the
+/// `SYSTEM_PROMPT_TOKEN_ESTIMATE` environment variable.
+fn system_prompt_token_estimate() -> i32 {
+    std::env::var("SYSTEM_PROMPT_TOKEN_ESTIMATE")
+        .ok()
+        .and_then(|v| v.parse::<i32>().ok())
+        .unwrap_or(5000)
+}

	let new_input = usage.usage.output_tokens.map(\|out\| out);
	let new_input = usage.usage.output_tokens;

Manual compaction counting fix + cli cleanup #5480

Manual compaction counting fix + cli cleanup #5480

Uh oh!

Conversation

katzdave commented Oct 30, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Copilot AI Nov 3, 2025

Choose a reason for hiding this comment

Uh oh!

jamadeo Nov 3, 2025

Choose a reason for hiding this comment

Uh oh!

katzdave Nov 3, 2025

Choose a reason for hiding this comment

Uh oh!

jamadeo Nov 3, 2025

Choose a reason for hiding this comment

Uh oh!

katzdave Nov 3, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Copilot AI Nov 3, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 3, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants