Claude 3.7 is out. we had some harcoded stuff #6197

DOsinga · 2025-12-19T16:36:55Z

Summary

removed refrences to 3.7 and simplified some code

katzdave · 2025-12-19T16:45:41Z

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

-    let re_version = Regex::new(r"-(\d+)-(\d+)-").unwrap();
-    if re_version.is_match(&result) {
-        result = re_version.replace(&result, "-$1.$2-").to_string();
+    let re_version_separator_pattern = Regex::new(r"-(\d+)-(\d+)-").unwrap();


map_to_canonical_model should be able to do this for you (although maybe not if the model ends up getting removed from openrouter)

katzdave · 2025-12-19T16:47:39Z

crates/goose/src/providers/formats/anthropic.rs

-    let max_tokens = model_config.max_tokens.unwrap_or(8192);
+    // https://platform.claude.com/docs/en/about-claude/models/overview
+    // 64k output tokens works for most claude models:
+    let max_tokens = model_config.max_tokens.unwrap_or(64 * 1024);


We should use max_completion_tokens from canonical model (and more generally across all providers). I can do a followup pass with that.

Seems like the data there correctly has 64k.

clouatre · 2025-12-19T16:53:17Z

Hi @DOsinga - I noticed our PRs both touch gcpvertexai.rs. PR #6187 addresses issue #6186 (global endpoint support for Gemini 3 models) and was submitted yesterday.

Your refactoring to simplify the model enums looks great! When you rebase to resolve conflicts, the key pieces from #6187 that need to stay are:

GcpLocation::Global variant with is_global() helper
build_host_url() method for global vs regional URL construction

These handle the case where Gemini 3 models need https://aiplatform.googleapis.com rather than https://global-aiplatform.googleapis.com.

Happy to help coordinate if needed!

Copilot

Pull request overview

This PR removes hardcoded references to Claude 3.7 models and simplifies the codebase to be more flexible with model versions. The changes include updating model lists, removing version-specific enums in GCP Vertex AI support, generalizing thinking/extended output logic, and updating tests.

Key Changes

Removed Claude 3.7 model references from known model lists across providers (Anthropic, OpenRouter)
Simplified GCP Vertex AI model handling by replacing version-specific enums with string-based model names
Generalized extended thinking support to work with all Claude models instead of being Claude 3.7-specific
Updated max output tokens default from 8192 to 64k for newer Claude models

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
ui/desktop/src/hooks/useCostTracking.ts	Minor formatting change consolidating function call to single line
ui/desktop/src/components/settings/app/AppSettingsSection.tsx	Removed extra blank line
crates/goose/src/providers/openrouter.rs	Updated known models list: removed claude-3.7-sonnet, added grok-code-fast-1, fixed gemini model name
crates/goose/src/providers/formats/gcpvertexai.rs	Major refactor: replaced version-specific enums with generic string-based model variants
crates/goose/src/providers/formats/databricks.rs	Removed outdated Claude 3.7-specific comments
crates/goose/src/providers/formats/anthropic.rs	Generalized thinking logic, removed temperature restriction, increased max_tokens default, removed Claude 3.7-specific test
crates/goose/src/providers/canonical/name_builder.rs	Simplified test cases while maintaining coverage
crates/goose/src/providers/anthropic.rs	Removed legacy Claude 3.7 models from known models list

Copilot · 2025-12-19T20:31:08Z

crates/goose/src/providers/formats/gcpvertexai.rs

    #[test]
-    fn test_generic_model_parsing() -> Result<()> {
-        // Test generic Claude models
+    fn test_model_parsing() -> Result<()> {


There are now two test functions with the same name test_model_parsing (lines 304 and 360). This will cause a compilation error. One of these tests should be removed or renamed.

Suggested change

fn test_model_parsing() -> Result<()> {

fn test_model_parsing_families() -> Result<()> {

Copilot · 2025-12-19T20:31:08Z

crates/goose/src/providers/formats/anthropic.rs

+            .insert("temperature".to_string(), json!(temp));
    }

    // Add thinking parameters for claude-3-7-sonnet model


The comment on line 430 still references "claude-3-7-sonnet model" but the check on line 432 now applies thinking parameters to ALL models when CLAUDE_THINKING_ENABLED is set, not just Claude 3.7. The comment should be updated to reflect this broader applicability.

Suggested change

// Add thinking parameters for claude-3-7-sonnet model

// Add thinking parameters for all models when CLAUDE_THINKING_ENABLED is set

Copilot · 2025-12-19T20:31:08Z

crates/goose/src/providers/formats/anthropic.rs

-    let max_tokens = model_config.max_tokens.unwrap_or(8192);
+    // https://platform.claude.com/docs/en/about-claude/models/overview
+    // 64k output tokens works for most claude models:
+    let max_tokens = model_config.max_tokens.unwrap_or(64 * 1024);


The default max_tokens was increased from 8192 to 65536 (64 * 1024). However, in the databricks.rs file at line 585, the default remains 8192. This inconsistency could lead to different behavior across providers for the same model when thinking is enabled.

Copilot

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 2 comments.

Copilot · 2026-01-05T15:40:52Z

crates/goose/src/providers/formats/anthropic.rs

    if let Some(temp) = model_config.temperature {
-        // Claude 3.7 models with thinking enabled don't support temperature
-        if !model_config.model_name.starts_with("claude-3-7-sonnet-") {
-            payload
-                .as_object_mut()
-                .unwrap()
-                .insert("temperature".to_string(), json!(temp));
-        }
+        payload
+            .as_object_mut()
+            .unwrap()
+            .insert("temperature".to_string(), json!(temp));


Temperature is now always added regardless of model when specified in model_config. The previous code excluded temperature for claude-3-7-sonnet models with thinking enabled. If users can still specify claude-3-7-sonnet models (even though they're removed from the known models list), this could cause API errors. Consider either: (1) explicitly blocking claude-3-7 model names, or (2) keeping the temperature exclusion for backward compatibility.

Copilot · 2026-01-05T15:40:53Z

crates/goose/src/providers/gcpvertexai.rs

-            GcpVertexAIModel::Gemini(GeminiVersion::Pro25Preview),
-            GcpVertexAIModel::Gemini(GeminiVersion::Flash25),
-            GcpVertexAIModel::Gemini(GeminiVersion::Pro25),
+            GcpVertexAIModel::Claude("claude-3-7-sonnet@20250219".to_string()),


Claude 3.7 model is still included in the known models list here, but the PR is removing Claude 3.7 references from other providers. This should be removed for consistency with the PR's goal.

Suggested change

GcpVertexAIModel::Claude("claude-3-7-sonnet@20250219".to_string()),

Copilot

Pull request overview

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

Copilot · 2026-01-06T20:04:06Z

crates/goose/src/providers/formats/anthropic.rs

+            .insert("temperature".to_string(), json!(temp));
    }

    // Add thinking parameters for claude-3-7-sonnet model


This comment still mentions "claude-3-7-sonnet model" but the thinking parameters are now applied to all models when CLAUDE_THINKING_ENABLED is set. Update the comment to reflect that thinking is no longer model-specific.

Suggested change

// Add thinking parameters for claude-3-7-sonnet model

// Add thinking parameters when CLAUDE_THINKING_ENABLED is set (applies to all models)

* 'main' of github.com:block/goose: Claude 3.7 is out. we had some harcoded stuff (#6197) Release 1.19.0 chore: upgrade to node v24 as engine (#6361) chore(deps): bump rsa from 0.9.9 to 0.9.10 (#6358) Bump rust toolchain to 1.92 (current stable) (#6356) Hide advanced recipe options under expandable content (#6021) fix: use .config/agents (plural) for skills directory (#6357) fix: prevent KaTeX from treating underscores as subscripts in plain text (#6242) fix: make goose review PRs more like goose contributors do (#6240) fix : preserve provider engine type when editing custom providers (#6106) feat(providers): add retry for model fetching (#6347) allow goose issue solver to react to activation comments (#6239)

* main: (31 commits) added validation and debug for invalid call tool result (#6368) Update MCP apps tutorial: fix _meta structure and version prereq (#6404) Fixed fonts (#6389) Update confidence levels prompt injection detection to reduce false positive rates (#6390) Add ML-based prompt injection detection (#5623) docs: update custom extensions tutorial (#6388) fix ResultsFormat error when loading old sessions (#6385) docs: add MCP Apps tutorial and documentation updates (#6384) changed z-index to make sure the search highlighter does not appear on modal overlay (#6386) Handling special claude model response in github copilot provider (#6369) fix: prevent duplicate rendering when tool returns both mcp-ui and mcp-apps resources (#6378) fix: update MCP Apps _meta.ui.resourceUri to use nested format (SEP-1865) (#6372) feat(providers): add streaming support for Google Gemini provider (#6191) Blog: edit links in mcp apps post (#6371) fix: prevent infinite loop of tool-input notifications in MCP Apps (#6374) fix: Show platform-specific keyboard shortcuts in UI (#6323) fix: we load extensions when agent starts so don't do it up front (#6350) docs: credit HumanLayer in RPI tutorial (#6365) Blog: Goose Lands MCP Apps (#6172) Claude 3.7 is out. we had some harcoded stuff (#6197) ...

Claude 3.7 is out. we had some harcoded stuff

78a8c55

DOsinga requested review from Copilot and katzdave and removed request for Copilot December 19, 2025 16:36

Copilot started reviewing on behalf of DOsinga December 19, 2025 16:37 View session

katzdave approved these changes Dec 19, 2025

View reviewed changes

Merge branch 'main' into depricate-claude-3.7

8f78b15

Copilot AI review requested due to automatic review settings December 19, 2025 20:28

Copilot started reviewing on behalf of DOsinga December 19, 2025 20:29 View session

Copilot AI reviewed Dec 19, 2025

View reviewed changes

Douwe Osinga added 3 commits January 5, 2026 10:21

Merge branch 'main' into depricate-claude-3.7

805a062

Fix

474755b

Cargo Fmt

d505ad6

Copilot AI review requested due to automatic review settings January 5, 2026 15:37

Copilot started reviewing on behalf of DOsinga January 5, 2026 15:37 View session

Copilot AI reviewed Jan 5, 2026

View reviewed changes

Douwe Osinga added 3 commits January 6, 2026 14:10

64K is 64000 according to Anthropic!

88a7112

Merge branch 'main' into depricate-claude-3.7

bf25242

Merge branch 'main' into depricate-claude-3.7

26c2627

Copilot AI review requested due to automatic review settings January 6, 2026 20:00

Copilot started reviewing on behalf of DOsinga January 6, 2026 20:01 View session

Copilot AI reviewed Jan 6, 2026

View reviewed changes

DOsinga merged commit 463eee3 into main Jan 6, 2026
26 checks passed

DOsinga deleted the depricate-claude-3.7 branch January 6, 2026 20:50

github-actions bot mentioned this pull request Jan 13, 2026

chore(release): release version 1.20.0 (minor) #6457

Merged

	fn test_model_parsing() -> Result<()> {
	fn test_model_parsing_families() -> Result<()> {

	// Add thinking parameters for claude-3-7-sonnet model
	// Add thinking parameters for all models when CLAUDE_THINKING_ENABLED is set

	// Add thinking parameters for claude-3-7-sonnet model
	// Add thinking parameters when CLAUDE_THINKING_ENABLED is set (applies to all models)

Claude 3.7 is out. we had some harcoded stuff #6197

Claude 3.7 is out. we had some harcoded stuff #6197

Uh oh!

Conversation

DOsinga commented Dec 19, 2025

Summary

Uh oh!

katzdave Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

katzdave Dec 19, 2025

Choose a reason for hiding this comment

Uh oh!

clouatre commented Dec 19, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Key Changes

Reviewed changes

Uh oh!

Copilot AI Dec 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 19, 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 Jan 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 5, 2026

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 Jan 6, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

katzdave Dec 19, 2025 •

edited

Loading