Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
110 changes: 95 additions & 15 deletions .roo/rules-release-notes-writer/1_main_workflow.xml
Original file line number Diff line number Diff line change
Expand Up @@ -294,9 +294,25 @@ fi
<parsing>
<regex>Detect PR references as: /#(\d{1,7})/g</regex>
<heuristics>
- If a bullet has no explicit PR number, attempt fuzzy matching to PR titles
- If ambiguous, mark as "unlinked" and exclude by default pending user choice
</heuristics>
Purpose: Count changelog bullets without explicit PR numbers as "referenced" when they confidently map to a PR.
Matching algorithm:
- Normalize both bullet text and PR titles:
- lowercase; remove punctuation; strip prefixes like "fix:", "feat:", "add:", "improve:", "chore:", "refactor:"
- remove parentheticals such as "(thanks @user!)", "(PR by @user)", "(#1234 ...)"
- collapse whitespace
- Tokenize and compute token-overlap score = |intersection(tokens)| / |union(tokens)|
- Author signal: if bullet contains "thanks @user", "by @user", or "PR by @user" and that user equals the PR author or credited issue reporter, add +0.20 to score
- Keyword boost: +0.05 when provider/model/domain keywords (e.g., OpenAI, Claude, Grok, Chutes, Qwen, LongCat, etc.) appear in both
Confidence thresholds:
- score ≥ 0.65 (after boosts) → linked (confident). Treat as changelog-referenced.
- 0.45 ≤ score &lt; 0.65 OR multiple candidates within 0.05 → ambiguous (needs review)
Copy link

Copilot AI Oct 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The condition 'multiple candidates within 0.05' is ambiguous. Clarify whether this means 'multiple candidates with scores within 0.05 of each other' or 'within 0.05 of the threshold'.

Suggested change
- 0.45 ≤ score &lt; 0.65 OR multiple candidates within 0.05 → ambiguous (needs review)
- 0.45 ≤ score &lt; 0.65 OR multiple candidates with scores within 0.05 of each other → ambiguous (needs review)

Copilot uses AI. Check for mistakes.
- score &lt; 0.45 → unlinked
Tie-breakers: higher score; if within 0.02 then same author; then closer merge date to release date; then lowest PR number
Edge case: If bullet credits exactly one username and exactly one PR in the window has that author, accept with score ≥ 0.50 (confidence="author-boost")
Implementation notes:
- Match only against PRs fetched for the version's date window
- Persist mapping bullet_text → { prNumber, confidenceScore, rationaleSignals[] } and use it to compute linked/ambiguous/unlinked counts
</heuristics>
</parsing>
</step>

Expand Down Expand Up @@ -649,9 +665,25 @@ fi
<parsing>
<regex>Detect PR references as: /#(\d{1,7})/g</regex>
<heuristics>
- If a bullet has no explicit PR number, attempt fuzzy matching to PR titles
- If ambiguous, mark as "unlinked" and exclude by default pending user choice
</heuristics>
Purpose: Count changelog bullets without explicit PR numbers as "referenced" when they confidently map to a PR.
Matching algorithm:
- Normalize both bullet text and PR titles:
- lowercase; remove punctuation; strip prefixes like "fix:", "feat:", "add:", "improve:", "chore:", "refactor:"
- remove parentheticals such as "(thanks @user!)", "(PR by @user)", "(#1234 ...)"
- collapse whitespace
- Tokenize and compute token-overlap score = |intersection(tokens)| / |union(tokens)|
- Author signal: if bullet contains "thanks @user", "by @user", or "PR by @user" and that user equals the PR author or credited issue reporter, add +0.20 to score
- Keyword boost: +0.05 when provider/model/domain keywords (e.g., OpenAI, Claude, Grok, Chutes, Qwen, LongCat, etc.) appear in both
Confidence thresholds:
- score ≥ 0.65 (after boosts) → linked (confident). Treat as changelog-referenced.
- 0.45 ≤ score &lt; 0.65 OR multiple candidates within 0.05 → ambiguous (needs review)
- score &lt; 0.45 → unlinked
Tie-breakers: higher score; if within 0.02 then same author; then closer merge date to release date; then lowest PR number
Edge case: If bullet credits exactly one username and exactly one PR in the window has that author, accept with score ≥ 0.50 (confidence="author-boost")
Implementation notes:
- Match only against PRs fetched for the version's date window
- Persist mapping bullet_text → { prNumber, confidenceScore, rationaleSignals[] } and use it to compute linked/ambiguous/unlinked counts
</heuristics>
</parsing>
</step>

Expand Down Expand Up @@ -993,9 +1025,25 @@ fi
<parsing>
<regex>Detect PR references as: /#(\d{1,7})/g</regex>
<heuristics>
- If a bullet has no explicit PR number, attempt fuzzy matching to PR titles
- If ambiguous, mark as "unlinked" and exclude by default pending user choice
</heuristics>
Purpose: Count changelog bullets without explicit PR numbers as "referenced" when they confidently map to a PR.
Matching algorithm:
- Normalize both bullet text and PR titles:
- lowercase; remove punctuation; strip prefixes like "fix:", "feat:", "add:", "improve:", "chore:", "refactor:"
- remove parentheticals such as "(thanks @user!)", "(PR by @user)", "(#1234 ...)"
- collapse whitespace
- Tokenize and compute token-overlap score = |intersection(tokens)| / |union(tokens)|
- Author signal: if bullet contains "thanks @user", "by @user", or "PR by @user" and that user equals the PR author or credited issue reporter, add +0.20 to score
- Keyword boost: +0.05 when provider/model/domain keywords (e.g., OpenAI, Claude, Grok, Chutes, Qwen, LongCat, etc.) appear in both
Confidence thresholds:
- score ≥ 0.65 (after boosts) → linked (confident). Treat as changelog-referenced.
- 0.45 ≤ score &lt; 0.65 OR multiple candidates within 0.05 → ambiguous (needs review)
- score &lt; 0.45 → unlinked
Tie-breakers: higher score; if within 0.02 then same author; then closer merge date to release date; then lowest PR number
Edge case: If bullet credits exactly one username and exactly one PR in the window has that author, accept with score ≥ 0.50 (confidence="author-boost")
Implementation notes:
- Match only against PRs fetched for the version's date window
- Persist mapping bullet_text → { prNumber, confidenceScore, rationaleSignals[] } and use it to compute linked/ambiguous/unlinked counts
</heuristics>
</parsing>
</step>

Expand Down Expand Up @@ -1335,9 +1383,25 @@ fi
<parsing>
<regex>Detect PR references as: /#(\d{1,7})/g</regex>
<heuristics>
- If a bullet has no explicit PR number, attempt fuzzy matching to PR titles
- If ambiguous, mark as "unlinked" and exclude by default pending user choice
</heuristics>
Purpose: Count changelog bullets without explicit PR numbers as "referenced" when they confidently map to a PR.
Matching algorithm:
- Normalize both bullet text and PR titles:
- lowercase; remove punctuation; strip prefixes like "fix:", "feat:", "add:", "improve:", "chore:", "refactor:"
- remove parentheticals such as "(thanks @user!)", "(PR by @user)", "(#1234 ...)"
- collapse whitespace
- Tokenize and compute token-overlap score = |intersection(tokens)| / |union(tokens)|
- Author signal: if bullet contains "thanks @user", "by @user", or "PR by @user" and that user equals the PR author or credited issue reporter, add +0.20 to score
- Keyword boost: +0.05 when provider/model/domain keywords (e.g., OpenAI, Claude, Grok, Chutes, Qwen, LongCat, etc.) appear in both
Confidence thresholds:
- score ≥ 0.65 (after boosts) → linked (confident). Treat as changelog-referenced.
- 0.45 ≤ score &lt; 0.65 OR multiple candidates within 0.05 → ambiguous (needs review)
- score &lt; 0.45 → unlinked
Tie-breakers: higher score; if within 0.02 then same author; then closer merge date to release date; then lowest PR number
Edge case: If bullet credits exactly one username and exactly one PR in the window has that author, accept with score ≥ 0.50 (confidence="author-boost")
Implementation notes:
- Match only against PRs fetched for the version's date window
- Persist mapping bullet_text → { prNumber, confidenceScore, rationaleSignals[] } and use it to compute linked/ambiguous/unlinked counts
</heuristics>
</parsing>
</step>

Expand Down Expand Up @@ -1687,9 +1751,25 @@ fi
<parsing>
<regex>Detect PR references as: /#(\d{1,7})/g</regex>
<heuristics>
- If a bullet has no explicit PR number, attempt fuzzy matching to PR titles
- If ambiguous, mark as "unlinked" and exclude by default pending user choice
</heuristics>
Purpose: Count changelog bullets without explicit PR numbers as "referenced" when they confidently map to a PR.
Matching algorithm:
- Normalize both bullet text and PR titles:
- lowercase; remove punctuation; strip prefixes like "fix:", "feat:", "add:", "improve:", "chore:", "refactor:"
- remove parentheticals such as "(thanks @user!)", "(PR by @user)", "(#1234 ...)"
- collapse whitespace
- Tokenize and compute token-overlap score = |intersection(tokens)| / |union(tokens)|
- Author signal: if bullet contains "thanks @user", "by @user", or "PR by @user" and that user equals the PR author or credited issue reporter, add +0.20 to score
- Keyword boost: +0.05 when provider/model/domain keywords (e.g., OpenAI, Claude, Grok, Chutes, Qwen, LongCat, etc.) appear in both
Confidence thresholds:
- score ≥ 0.65 (after boosts) → linked (confident). Treat as changelog-referenced.
- 0.45 ≤ score &lt; 0.65 OR multiple candidates within 0.05 → ambiguous (needs review)
- score &lt; 0.45 → unlinked
Tie-breakers: higher score; if within 0.02 then same author; then closer merge date to release date; then lowest PR number
Edge case: If bullet credits exactly one username and exactly one PR in the window has that author, accept with score ≥ 0.50 (confidence="author-boost")
Implementation notes:
- Match only against PRs fetched for the version's date window
- Persist mapping bullet_text → { prNumber, confidenceScore, rationaleSignals[] } and use it to compute linked/ambiguous/unlinked counts
</heuristics>
</parsing>
</step>

Expand Down
1 change: 1 addition & 0 deletions docs/update-notes/index.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@ image: /img/social-share.jpg

### Version 3.29

* [3.29.2](/update-notes/v3.29.2) (2025-10-27)
* [3.29.1](/update-notes/v3.29.1) (2025-10-27)
* [3.29.0](/update-notes/v3.29.0) (2025-10-24)

Expand Down
30 changes: 30 additions & 0 deletions docs/update-notes/v3.29.2.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
---
description: Roo Code 3.29.2 adds higher-context Chutes AI models, improves settings text, and fixes provider caching and free-model pricing display.
keywords:
- roo code 3.29.2
- chutes ai
- longcat
- provider update
- bug fixes
- free models
- pricing display
image: /img/social-share.jpg
---

# Roo Code 3.29.2 Release Notes (2025-10-27)

This patch adds higher-context Chutes AI models, cleans up settings text, and fixes provider caching and free‑model pricing display.


## Provider Updates

* Chutes AI: Add LongCat‑Flash‑Thinking‑FP8 models (200K, 128K) for longer coding sessions with faster, cost‑effective performance (thanks leakless21!) ([#8426](https://github.com/RooCodeInc/Roo-Code/pull/8426))
Comment on lines +16 to +21
Copy link

Copilot AI Oct 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Inconsistent use of non-breaking hyphens (‑) in 'LongCat‑Flash‑Thinking‑FP8' and 'cost‑effective'. These should use standard hyphens (-) for consistency with standard markdown formatting.

Suggested change
This patch adds higher-context Chutes AI models, cleans up settings text, and fixes provider caching and freemodel pricing display.
## Provider Updates
* Chutes AI: Add LongCatFlashThinkingFP8 models (200K, 128K) for longer coding sessions with faster, costeffective performance (thanks leakless21!) ([#8426](https://github.com/RooCodeInc/Roo-Code/pull/8426))
This patch adds higher-context Chutes AI models, cleans up settings text, and fixes provider caching and free-model pricing display.
## Provider Updates
* Chutes AI: Add LongCat-Flash-Thinking-FP8 models (200K, 128K) for longer coding sessions with faster, cost-effective performance (thanks leakless21!) ([#8426](https://github.com/RooCodeInc/Roo-Code/pull/8426))

Copilot uses AI. Check for mistakes.

## QOL Improvements

* Settings descriptions: Remove specific "Claude 3.7 Sonnet" version across locales to keep guidance current (thanks rwydaegh!) ([#8437](https://github.com/RooCodeInc/Roo-Code/pull/8437))

## Bug Fixes

* Roo provider: Fix caching logic to emit correct usage data and apply proper protocol ([#8860](https://github.com/RooCodeInc/Roo-Code/pull/8860))
* Free models: Hide pricing info and show zero cost to avoid confusion ([#8864](https://github.com/RooCodeInc/Roo-Code/pull/8864))
1 change: 1 addition & 0 deletions sidebars.ts
Original file line number Diff line number Diff line change
Expand Up @@ -222,6 +222,7 @@ const sidebars: SidebarsConfig = {
type: 'category',
label: '3.29',
items: [
{ type: 'doc', id: 'update-notes/v3.29.2', label: '3.29.2' },
{ type: 'doc', id: 'update-notes/v3.29.1', label: '3.29.1' },
{ type: 'doc', id: 'update-notes/v3.29.0', label: '3.29.0' },
],
Expand Down
Loading