feat(sessions): add Command Code as a tracked source by baanish · Pull Request #717 · junhoyeo/tokscale

baanish · 2026-06-16T20:00:19Z

What

Adds Command Code (commandcode.ai) as a tracked session source, scanning ~/.commandcode/projects/<slug>/*.jsonl.

Why it's estimated

Command Code does not persist token usage on disk — the CLI computes usage in memory and ships it to its backend (confirmed by decompiling the npm bundle; the local transcript carries only message text). So tokens are estimated from the transcript, the same way tokscale already handles kiro:

input = cumulative conversation context preceding each assistant turn (Command Code re-sends the full context per request)
output = that turn's own content
~4 chars/token, counted from each message's canonical JSON serialization so tool args/results are included

The model id comes from ~/.commandcode/config.json and is canonicalized (e.g. MiniMaxAI/MiniMax-M3-Free → MiniMax-M3) so pricing resolves to the real paid model instead of the free-promo entry or a fuzzy mismatch. The source bypasses the parse cache because that model dependency lives outside the fingerprinted transcript file.

Known limitation

The estimate approximates tokens processed and will not match Command Code's server dashboard: it counts full tool outputs that the tool truncates before billing, and cannot see auxiliary (tool-desc / taste-1) runs. On a real ~880-turn session it estimates 270M input / ~$81 at the MiniMax-M3 rate, versus the dashboard's 181M / $0.00 (a promo). This is consistent with how tokscale treats its other estimated sources.

Testing

9 parser unit tests (estimation, model canonicalization, checkpoint/malformed/empty edge cases)
registry + CLI mirror-point tests updated (ClientId::COUNT, ClientFilter, TUI client list)
real-session smoke via tokscale clients and tokscale models -c commandcode
full suites green: core 927, CLI 778

Reviewed adversarially with Codex (2 rounds); fixes for model-pricing resolution and config-aware (cache-bypass) parsing folded in.

🤖 Generated with Claude Code

Related to #668

Summary by cubic

Add commandcode as a tracked session source with estimated token usage and pricing so Command Code sessions show up with costs and analytics. Scans local transcripts and applies model-aware pricing without relying on server usage; registered as client id 29.

New Features
- Scan ~/.commandcode/projects/<slug>/*.jsonl; skip *.checkpoints.jsonl.
- Estimate tokens from canonical JSON (~4 chars/token): input = cumulative context before each assistant turn; output = that turn’s content; cache_read = 0.
- Read model from ~/.commandcode/config.json, canonicalize for pricing (e.g., MiniMaxAI/MiniMax-M3-Free → MiniMax-M3), and bypass the parse cache so config changes take effect immediately.
- Register across the app: CLI filter flag --commandcode and name mapping, TUI entry “Command Code” with hotkey d, client id 29 (ClientId::COUNT = 30), and frontend display name/logo/color.

^{Written for commit 9c1bcd1. Summary will update on new commits.}

vercel · 2026-06-16T20:00:28Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
tokscale	Ready	Preview, Comment	Jun 17, 2026 6:42pm

cubic-dev-ai

No issues found across 9 files

_{Re-trigger cubic}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: aab8e1edfe

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-06-16T20:05:41Z

+
+        let chars = entry.content.as_ref().map(content_chars).unwrap_or(0);
+
+        match entry.role.as_deref() {


Parse Command Code records by type

When scanning current Command Code JSONL sessions, the records use top-level type values (user/assistant) rather than role, so entry.role is None and every line falls into the catch-all arm. That means no assistant UnifiedMessages are ever emitted and tokscale models -c commandcode reports zero usage despite discovering session files; read type (and the corresponding message content shape) instead of matching role here.

Useful? React with 👍 / 👎.

Note: this reply is posted by Claude Code on behalf of the author, not written by them.

This finding doesn't apply to Command Code's format. Command Code transcripts use a top-level role field (user / assistant / tool), not a top-level type. The type field only appears inside content parts (e.g. {"type":"text", ...}), which the match entry.role here never touches.

Verified against a real session (1,766 lines): every line carries role (880 assistant, 862 tool, 24 user), and zero lines have a top-level type. Running it end to end, tokscale models -c commandcode reports ~270M input across 880 messages — not zero usage.

The suggestion looks like it assumed the OpenAI/Codex transcript shape, which does use a top-level type.

Command Code (commandcode.ai) stores session transcripts under ~/.commandcode/projects but does not persist token usage on disk — the CLI computes usage in memory and sends it to its backend. Estimate usage from the transcript: input from the cumulative conversation context preceding each assistant turn and output from the assistant's own content, at ~4 characters per token (matching tokscale's other estimated sources), counted from each message's canonical JSON serialization so structured tool args/results are included. Canonicalize the configured model id from ~/.commandcode/config.json (e.g. "MiniMaxAI/MiniMax-M3-Free" -> "MiniMax-M3") so pricing resolves to the real paid model rather than the free-promo entry or a fuzzy mismatch. The estimate approximates tokens processed and intentionally does not match Command Code's server-reported usage (which reflects tool-output truncation and auxiliary model runs absent from the transcript). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Directive: input estimation is intentionally an upper bound; changing re-sent context to cache_read is a maintainer decision needing real billing data Confidence: high Scope-risk: narrow

cubic-dev-ai

1 issue found across 1 file (changes from recent commits).

Prompt for AI agents (unresolved issues)


Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="crates/tokscale-core/src/sessions/commandcode.rs">

<violation number="1" location="crates/tokscale-core/src/sessions/commandcode.rs:411">
P2: Test claiming to pin exact token values derives expectations from the same helpers as production, defeating its purpose</violation>
</file>

_{Reply with feedback, questions, or to request a fix.

Re-trigger cubic}

cubic-dev-ai · 2026-06-17T18:09:59Z

+        let user2_content = json!([{"type": "text", "text": "cccc"}]);
+        let asst2_content = json!([{"type": "text", "text": "dddd"}]);
+
+        let user1_chars = content_chars(&user1_content);


P2: Test claiming to pin exact token values derives expectations from the same helpers as production, defeating its purpose

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At crates/tokscale-core/src/sessions/commandcode.rs, line 411: <comment>Test claiming to pin exact token values derives expectations from the same helpers as production, defeating its purpose</comment> <file context> @@ -355,6 +368,108 @@ mod tests { + let user2_content = json!([{"type": "text", "text": "cccc"}]); + let asst2_content = json!([{"type": "text", "text": "dddd"}]); + + let user1_chars = content_chars(&user1_content); + let asst1_chars = content_chars(&asst1_content); + let user2_chars = content_chars(&user2_content); </file context>

…t cmd, Warp/Cline (#720) * docs(readme): sync EN/ja/ko/zh — add Command Code, Antigravity CLI, report cmd, Warp/Cline EN is the canonical lead. Brings the localized READMEs into sync: - Add Command Code (#717) and Antigravity CLI (#713) to the client support table, multi-platform feature list, source-filtering list, and the `--client` possible-values list (all four locales). - Document the `tokscale report` command (summarize-on-by-default, apple-fm backend, --no-summarize opt-out, wiki DB, task grouping) in ja/ko/zh. - Back-fill pre-existing translation drift: Warp/Oz + Cline client rows, feature-list entries, and the Warp/Oz Commands section in ja/ko/zh; fix a malformed (logo-less) Grok Build row in ko/zh. All four files now share an identical 33-client roster and --client list. Constraint: README.md is the lead; localized files mirror it Confidence: high Scope-risk: narrow Not-tested: native-speaker review of ja/ko/zh translation nuance * docs(readme): address Codex/cubic review on the report summarizer docs Apply review feedback from PR #720 (Codex + cubic), verified against the code, across all four locales (EN lead): - apple-fm prerequisite (Codex/cubic P2): scripts/wiki-summarizer.py is NOT bundled in the published npm/bunx packages (files lists ship only bin/dist), so the documented default backend fails with "wiki-summarizer.py not found". Add a note that npm/bunx users must copy the script into the config dir or use a CLI backend (--summarizer claude/codex/gemini/kiro). - Apple FM fallback (cubic P3): when the Apple FM SDK is unavailable the script falls back to a heuristic classifier (fallback_classify), not "empty summaries" — corrected the wording in all locales. - wiki.db path (cubic P3): WikiDb::default_path() uses dirs::config_dir(), so it is platform-specific; qualified the path instead of hardcoding ~/.config. Confidence: high Scope-risk: narrow

…turn input (#750) Three confirmed bugs in the Command Code session parser (PR #717): (a) PANIC: canonicalize_model byte-sliced `base[base.len()-N..]` guarded only by a length check. A non-ASCII model id from the untrusted ~/.commandcode/config.json could place that byte index mid-codepoint and panic. Switch to a boundary-safe `str::get(..)` slice that returns None (instead of panicking) when the tail is not on a char boundary, preserving the case-insensitive `-free` strip. (b) PRICING: dropping the org prefix turned `MiniMaxAI/MiniMax-M3-Free` into bare `MiniMax-M3`, but the client's own `command-code` provider is not a pricing provider, so the resolver never reached a `minimax/...` key and pricing was lost. Recover the real provider from the gateway id via `provider_identity::inferred_provider_from_model` and use it as the message provider_id (matching the claudecode convention), falling back to `command-code` when nothing is inferred. (c) O(N^2) INPUT: per-assistant-turn input was estimated from `context_chars`, a running cumulative sum of ALL prior messages, inflating input across the session versus other clients. Estimate per-turn input from only the new context introduced since the previous response (user prompt + tool results). Summed over the session this charges each message's content once. Regression tests added: non-ASCII / emoji model ids do not panic; a MiniMax model resolves non-zero pricing; per-turn input does not accumulate. Updated the module doc-comment and renamed the pinning test to match the new model. Confidence: high Scope-risk: moderate Directive: provider_id for MiniMax sessions is now `minimax`, not `command-code`; this is intentional so pricing resolves. Do not revert without restoring a pricing path for the gateway provider. Not-tested: real-world non-MiniMax gateway ids that infer no provider (fall back to `command-code`, unchanged behavior)

…746) * docs: sync EN/ja/ko/zh-cn for unreleased clients + breaking flag removal Pre-v3.2.0 documentation sync for the unreleased range (v3.1.3..main): - #465: replace the now-false legacy per-client flag notices in ja/ko/zh-cn (they claimed the removed flags still work) with a v3.2.0 breaking-change migration note; add the same note to README.md (which had none). - #728: document the MiniMax Token Plan subscription source (distinct from the MINIMAX_API_KEY row) in all locales; port the entire Subscription Usage section into ja/ko/zh-cn (was English-only). - #718: add the Jcode table row + detail section to the locales missing them. - #726: document the TOKSCALE_FM_DEBUG env var in all locales. - #633: add the missing task-attributed report bullet to README.ja Key Features. - drift: add Junie to the frontend Source-filtering list (all locales). - #710: fix the MiMo Code repo link (XiaomiMiMo/MiMo -> XiaomiMiMo/MiMo-Code). - #717: disclose Command Code token usage is estimated (~4 chars/token). Confidence: medium Scope-risk: narrow Directive: ja/ko/zh-cn translations of the ported Subscription Usage section are machine-generated and should get a native-speaker review pass Not-tested: #713 Antigravity CLI detail section was not added — no English source section exists to port from * fix(report): feed real session content to the summarizer (#633) extract_content_for_session unconditionally returned metadata_only_content() (first_user_message hardcoded None), so the report summarizer never saw any conversation content and the four real per-client extractors were dead code. Add content_extractor::extract_session_content, which dispatches to the correct per-client extractor (opencode/claude/codex/gemini) and falls back to metadata-only — never erroring or panicking — for unknown clients, missing candidates, or unreadable/unparseable files. report.rs builds a SessionPathIndex once (session_id -> transcript file, plus opencode DBs) and threads it through run_summarizer so each payload carries the real first user message. Confidence: high Scope-risk: moderate Rejected: thread file paths through core's scanner/WikiEntry | too invasive; indexed at the report layer instead Not-tested: end-to-end opencode/codex/gemini extraction in report.rs (core dispatcher covers claude + all fallback paths; per-client extractors are pre-existing) * docs: name the breaking release v4.0.0 (was v3.2.0) The per-client flag removal (#465) is a breaking change, so the next release is v4.0.0, not v3.2.0. Update the migration notes in all four README locales and the main.rs doc comment accordingly. * fix(report): real Codex/Gemini extraction + (client,session_id) index keying Addresses automated review feedback on the #633 report-summarizer-content fix (PR #746). The summarizer still surfaced (none) for normal Codex/Gemini sessions and could mis-route cross-client session_id collisions. - content_extractor: parse the current on-disk Codex format (event_msg with payload.type == "user_message", text in payload.message) and skip harness-injected context blocks (<environment_context>/<system-reminder>/ <user_instructions>), mirroring sessions::codex. - content_extractor: Gemini extractor now handles chat-recording JSON (messages[].type == "user" / content) and falls back to scanning line-delimited JSONL; empty/whitespace user text is treated as not-found. - extract_session_content: an empty/whitespace first_user_message no longer counts as success, so scanning continues to a later candidate with real text. - report: SessionPathIndex is keyed by (client, session_id) to prevent cross-client collisions, and Gemini files are keyed by their in-file sessionId (via gemini_session_id_for_file) rather than the filename stem, since the wiki entry's session_id is derived from inside the file. - Added fixture-based regression tests for all of the above. Constraint: wiki session_id for Gemini comes from the in-file sessionId, not the path stem Rejected: match any leading '<' for Codex injected blocks | drops legit prompts starting with markup Confidence: high Scope-risk: narrow

* feat(sessions): add Command Code as a tracked source Command Code (commandcode.ai) stores session transcripts under ~/.commandcode/projects but does not persist token usage on disk — the CLI computes usage in memory and sends it to its backend. Estimate usage from the transcript: input from the cumulative conversation context preceding each assistant turn and output from the assistant's own content, at ~4 characters per token (matching tokscale's other estimated sources), counted from each message's canonical JSON serialization so structured tool args/results are included. Canonicalize the configured model id from ~/.commandcode/config.json (e.g. "MiniMaxAI/MiniMax-M3-Free" -> "MiniMax-M3") so pricing resolves to the real paid model rather than the free-promo entry or a fuzzy mismatch. The estimate approximates tokens processed and intentionally does not match Command Code's server-reported usage (which reflects tool-output truncation and auxiliary model runs absent from the transcript). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * test(commandcode): document and pin cumulative-input estimation behavior Directive: input estimation is intentionally an upper bound; changing re-sent context to cache_read is a maintainer decision needing real billing data Confidence: high Scope-risk: narrow * style: apply rustfmt --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> Co-authored-by: Junho Yeo <i@junho.io>

…t cmd, Warp/Cline (junhoyeo#720) * docs(readme): sync EN/ja/ko/zh — add Command Code, Antigravity CLI, report cmd, Warp/Cline EN is the canonical lead. Brings the localized READMEs into sync: - Add Command Code (junhoyeo#717) and Antigravity CLI (junhoyeo#713) to the client support table, multi-platform feature list, source-filtering list, and the `--client` possible-values list (all four locales). - Document the `tokscale report` command (summarize-on-by-default, apple-fm backend, --no-summarize opt-out, wiki DB, task grouping) in ja/ko/zh. - Back-fill pre-existing translation drift: Warp/Oz + Cline client rows, feature-list entries, and the Warp/Oz Commands section in ja/ko/zh; fix a malformed (logo-less) Grok Build row in ko/zh. All four files now share an identical 33-client roster and --client list. Constraint: README.md is the lead; localized files mirror it Confidence: high Scope-risk: narrow Not-tested: native-speaker review of ja/ko/zh translation nuance * docs(readme): address Codex/cubic review on the report summarizer docs Apply review feedback from PR junhoyeo#720 (Codex + cubic), verified against the code, across all four locales (EN lead): - apple-fm prerequisite (Codex/cubic P2): scripts/wiki-summarizer.py is NOT bundled in the published npm/bunx packages (files lists ship only bin/dist), so the documented default backend fails with "wiki-summarizer.py not found". Add a note that npm/bunx users must copy the script into the config dir or use a CLI backend (--summarizer claude/codex/gemini/kiro). - Apple FM fallback (cubic P3): when the Apple FM SDK is unavailable the script falls back to a heuristic classifier (fallback_classify), not "empty summaries" — corrected the wording in all locales. - wiki.db path (cubic P3): WikiDb::default_path() uses dirs::config_dir(), so it is platform-specific; qualified the path instead of hardcoding ~/.config. Confidence: high Scope-risk: narrow

* feat(sessions): add Command Code as a tracked source Command Code (commandcode.ai) stores session transcripts under ~/.commandcode/projects but does not persist token usage on disk — the CLI computes usage in memory and sends it to its backend. Estimate usage from the transcript: input from the cumulative conversation context preceding each assistant turn and output from the assistant's own content, at ~4 characters per token (matching tokscale's other estimated sources), counted from each message's canonical JSON serialization so structured tool args/results are included. Canonicalize the configured model id from ~/.commandcode/config.json (e.g. "MiniMaxAI/MiniMax-M3-Free" -> "MiniMax-M3") so pricing resolves to the real paid model rather than the free-promo entry or a fuzzy mismatch. The estimate approximates tokens processed and intentionally does not match Command Code's server-reported usage (which reflects tool-output truncation and auxiliary model runs absent from the transcript). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * test(commandcode): document and pin cumulative-input estimation behavior Directive: input estimation is intentionally an upper bound; changing re-sent context to cache_read is a maintainer decision needing real billing data Confidence: high Scope-risk: narrow * style: apply rustfmt --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> Co-authored-by: Junho Yeo <i@junho.io>

…t cmd, Warp/Cline (junhoyeo#720) * docs(readme): sync EN/ja/ko/zh — add Command Code, Antigravity CLI, report cmd, Warp/Cline EN is the canonical lead. Brings the localized READMEs into sync: - Add Command Code (junhoyeo#717) and Antigravity CLI (junhoyeo#713) to the client support table, multi-platform feature list, source-filtering list, and the `--client` possible-values list (all four locales). - Document the `tokscale report` command (summarize-on-by-default, apple-fm backend, --no-summarize opt-out, wiki DB, task grouping) in ja/ko/zh. - Back-fill pre-existing translation drift: Warp/Oz + Cline client rows, feature-list entries, and the Warp/Oz Commands section in ja/ko/zh; fix a malformed (logo-less) Grok Build row in ko/zh. All four files now share an identical 33-client roster and --client list. Constraint: README.md is the lead; localized files mirror it Confidence: high Scope-risk: narrow Not-tested: native-speaker review of ja/ko/zh translation nuance * docs(readme): address Codex/cubic review on the report summarizer docs Apply review feedback from PR junhoyeo#720 (Codex + cubic), verified against the code, across all four locales (EN lead): - apple-fm prerequisite (Codex/cubic P2): scripts/wiki-summarizer.py is NOT bundled in the published npm/bunx packages (files lists ship only bin/dist), so the documented default backend fails with "wiki-summarizer.py not found". Add a note that npm/bunx users must copy the script into the config dir or use a CLI backend (--summarizer claude/codex/gemini/kiro). - Apple FM fallback (cubic P3): when the Apple FM SDK is unavailable the script falls back to a heuristic classifier (fallback_classify), not "empty summaries" — corrected the wording in all locales. - wiki.db path (cubic P3): WikiDb::default_path() uses dirs::config_dir(), so it is platform-specific; qualified the path instead of hardcoding ~/.config. Confidence: high Scope-risk: narrow

…turn input (junhoyeo#750) Three confirmed bugs in the Command Code session parser (PR junhoyeo#717): (a) PANIC: canonicalize_model byte-sliced `base[base.len()-N..]` guarded only by a length check. A non-ASCII model id from the untrusted ~/.commandcode/config.json could place that byte index mid-codepoint and panic. Switch to a boundary-safe `str::get(..)` slice that returns None (instead of panicking) when the tail is not on a char boundary, preserving the case-insensitive `-free` strip. (b) PRICING: dropping the org prefix turned `MiniMaxAI/MiniMax-M3-Free` into bare `MiniMax-M3`, but the client's own `command-code` provider is not a pricing provider, so the resolver never reached a `minimax/...` key and pricing was lost. Recover the real provider from the gateway id via `provider_identity::inferred_provider_from_model` and use it as the message provider_id (matching the claudecode convention), falling back to `command-code` when nothing is inferred. (c) O(N^2) INPUT: per-assistant-turn input was estimated from `context_chars`, a running cumulative sum of ALL prior messages, inflating input across the session versus other clients. Estimate per-turn input from only the new context introduced since the previous response (user prompt + tool results). Summed over the session this charges each message's content once. Regression tests added: non-ASCII / emoji model ids do not panic; a MiniMax model resolves non-zero pricing; per-turn input does not accumulate. Updated the module doc-comment and renamed the pinning test to match the new model. Confidence: high Scope-risk: moderate Directive: provider_id for MiniMax sessions is now `minimax`, not `command-code`; this is intentional so pricing resolves. Do not revert without restoring a pricing path for the gateway provider. Not-tested: real-world non-MiniMax gateway ids that infer no provider (fall back to `command-code`, unchanged behavior)

…unhoyeo#746) * docs: sync EN/ja/ko/zh-cn for unreleased clients + breaking flag removal Pre-v3.2.0 documentation sync for the unreleased range (v3.1.3..main): - junhoyeo#465: replace the now-false legacy per-client flag notices in ja/ko/zh-cn (they claimed the removed flags still work) with a v3.2.0 breaking-change migration note; add the same note to README.md (which had none). - junhoyeo#728: document the MiniMax Token Plan subscription source (distinct from the MINIMAX_API_KEY row) in all locales; port the entire Subscription Usage section into ja/ko/zh-cn (was English-only). - junhoyeo#718: add the Jcode table row + detail section to the locales missing them. - junhoyeo#726: document the TOKSCALE_FM_DEBUG env var in all locales. - junhoyeo#633: add the missing task-attributed report bullet to README.ja Key Features. - drift: add Junie to the frontend Source-filtering list (all locales). - junhoyeo#710: fix the MiMo Code repo link (XiaomiMiMo/MiMo -> XiaomiMiMo/MiMo-Code). - junhoyeo#717: disclose Command Code token usage is estimated (~4 chars/token). Confidence: medium Scope-risk: narrow Directive: ja/ko/zh-cn translations of the ported Subscription Usage section are machine-generated and should get a native-speaker review pass Not-tested: junhoyeo#713 Antigravity CLI detail section was not added — no English source section exists to port from * fix(report): feed real session content to the summarizer (junhoyeo#633) extract_content_for_session unconditionally returned metadata_only_content() (first_user_message hardcoded None), so the report summarizer never saw any conversation content and the four real per-client extractors were dead code. Add content_extractor::extract_session_content, which dispatches to the correct per-client extractor (opencode/claude/codex/gemini) and falls back to metadata-only — never erroring or panicking — for unknown clients, missing candidates, or unreadable/unparseable files. report.rs builds a SessionPathIndex once (session_id -> transcript file, plus opencode DBs) and threads it through run_summarizer so each payload carries the real first user message. Confidence: high Scope-risk: moderate Rejected: thread file paths through core's scanner/WikiEntry | too invasive; indexed at the report layer instead Not-tested: end-to-end opencode/codex/gemini extraction in report.rs (core dispatcher covers claude + all fallback paths; per-client extractors are pre-existing) * docs: name the breaking release v4.0.0 (was v3.2.0) The per-client flag removal (junhoyeo#465) is a breaking change, so the next release is v4.0.0, not v3.2.0. Update the migration notes in all four README locales and the main.rs doc comment accordingly. * fix(report): real Codex/Gemini extraction + (client,session_id) index keying Addresses automated review feedback on the junhoyeo#633 report-summarizer-content fix (PR junhoyeo#746). The summarizer still surfaced (none) for normal Codex/Gemini sessions and could mis-route cross-client session_id collisions. - content_extractor: parse the current on-disk Codex format (event_msg with payload.type == "user_message", text in payload.message) and skip harness-injected context blocks (<environment_context>/<system-reminder>/ <user_instructions>), mirroring sessions::codex. - content_extractor: Gemini extractor now handles chat-recording JSON (messages[].type == "user" / content) and falls back to scanning line-delimited JSONL; empty/whitespace user text is treated as not-found. - extract_session_content: an empty/whitespace first_user_message no longer counts as success, so scanning continues to a later candidate with real text. - report: SessionPathIndex is keyed by (client, session_id) to prevent cross-client collisions, and Gemini files are keyed by their in-file sessionId (via gemini_session_id_for_file) rather than the filename stem, since the wiki entry's session_id is derived from inside the file. - Added fixture-based regression tests for all of the above. Constraint: wiki session_id for Gemini comes from the in-file sessionId, not the path stem Rejected: match any leading '<' for Codex injected blocks | drops legit prompts starting with markup Confidence: high Scope-risk: narrow

vercel Bot deployed to Preview June 16, 2026 20:01 View deployment

cubic-dev-ai Bot reviewed Jun 16, 2026

View reviewed changes

chatgpt-codex-connector Bot reviewed Jun 16, 2026

View reviewed changes

baanish force-pushed the main branch from aab8e1e to c52436d Compare June 16, 2026 20:26

vercel Bot deployed to Preview June 16, 2026 20:27 View deployment

baanish mentioned this pull request Jun 17, 2026

feat: support Command Code (commandcode.ai) usage tracking via API sync #668

Open

test(commandcode): document and pin cumulative-input estimation behavior

44e84c8

Directive: input estimation is intentionally an upper bound; changing re-sent context to cache_read is a maintainer decision needing real billing data Confidence: high Scope-risk: narrow

cubic-dev-ai Bot reviewed Jun 17, 2026

View reviewed changes

junhoyeo added 2 commits June 18, 2026 03:28

style: apply rustfmt

6279de4

Merge origin/main; renumber CommandCode to 29

9c1bcd1

junhoyeo merged commit 18c7e87 into junhoyeo:main Jun 17, 2026
2 checks passed

vercel Bot deployed to Preview June 17, 2026 18:42 View deployment

junhoyeo mentioned this pull request Jun 17, 2026

docs(readme): sync EN/ja/ko/zh — Command Code, Antigravity CLI, report cmd, Warp/Cline #720

Merged

This was referenced Jun 22, 2026

chore(release): pre-v4.0.0 doc sync + fix report summarizer content #746

Merged

fix(commandcode): char-safe model canonicalize, MiniMax pricing, per-turn input #750

Merged

astkaasa mentioned this pull request Jun 22, 2026

feat: support commandcode.ai #508

Closed

junhoyeo mentioned this pull request Jul 6, 2026

Will it support other AI tools in the future? #450

Closed

Navigation Menu

feat(sessions): add Command Code as a tracked source - #717

feat(sessions): add Command Code as a tracked source#717
junhoyeo merged 4 commits into
junhoyeo:mainfrom
baanish:main

baanish commented Jun 16, 2026 •

edited by cubic-dev-ai Bot

Loading

Uh oh!

vercel Bot commented Jun 16, 2026 •

edited

Loading

Uh oh!

cubic-dev-ai Bot left a comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jun 16, 2026

Uh oh!

baanish Jun 16, 2026

Uh oh!

cubic-dev-ai Bot left a comment

Uh oh!

cubic-dev-ai Bot Jun 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants


		let chars = entry.content.as_ref().map(content_chars).unwrap_or(0);

		match entry.role.as_deref() {

Conversation

baanish commented Jun 16, 2026 • edited by cubic-dev-ai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Why it's estimated

Known limitation

Testing

Summary by cubic

Uh oh!

vercel Bot commented Jun 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cubic-dev-ai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jun 16, 2026

Choose a reason for hiding this comment

Uh oh!

baanish Jun 16, 2026

Choose a reason for hiding this comment

Uh oh!

cubic-dev-ai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

cubic-dev-ai Bot Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

baanish commented Jun 16, 2026 •

edited by cubic-dev-ai Bot

Loading

vercel Bot commented Jun 16, 2026 •

edited

Loading