Repository navigation
v0.11.2: restore GPT-6.1-Sol reduction and repair concurrent TT discovery - #21
Merged
Merged
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Owner-authorized repair and release
This addresses the two release blockers left after #20: GPT-6.1-Sol exited the ContextEngine before JEV because the upstream tokenizer-name table lacked the model, and concurrent Hermes directory discovery could consume TT before initialization finished.
GPT-6.1-Sol
Add an explicit local
gpt-6.1-sol -> o200k_basecompatibility entry, including known OpenAI/Codex/OpenRouter-qualified names. Upstream tiktoken mappings and explicit operator configuration retain precedence. No blanket GPT-6/future-version assumption; unrelated namespaces cannot inherit compatibility via the cache.Actual provider check: five synthetic inputs were sent to the documented OpenAI
/v1/responses/input_tokensendpoint, not to generation. English, multilingual/emoji text, code/numbers, JSON and literal special-token-looking strings all matched o200k_base content counts plus the same six-token message framing. Three cases distinguish this encoding from cl100k_base. The exact synthetic inputs/results are committed intests/fixtures/gpt61-token-counts.json. This is empirical compatibility evidence, not an upstream model declaration or universal proof. Runtime remains local, with no new API key or remote count dependency.The new pipeline test uses the actual Sol model ID with no manual encoding override, proves JEV is reached, verifies complete-request token reduction, preserves system/current wording, and reads omitted evidence back exactly. Canonical request JSON remains the measurement boundary, not hidden provider billing framing.
Concurrent discovery / multiplex profiles
The race happens before TT's module can execute, so a plugin-local registration fix cannot synchronize every reader. The explicit
install-context-engineCLI now requests a narrow, source-hash-guarded repair to the reviewed affected Hermesplugins/plugin_loader.py.Only TT user ContextEngine loads are wrapped. Initialization is synchronized per resolved profile path, profile caches use distinct namespaces, recursive same-thread discovery never exposes a partial module, and failed loads can retry. Unrelated plugins keep the original loader. An adjacent byte-exact backup is retained; unknown source versions, modified patches, symlinks, mismatching backups and concurrent edits are refused. The ordinary Python installer API defaults to no host patch;
install-dashboardand package import do not patch the host.This host-file change is explicitly documented in README, migration, security and installation guides. It is not disguised as a core-free repair. No conversation, vault, credentials or selected engine configuration is rewritten.
Tests completed before push
188a1a5loader source: 9 passed, only temporary files modified.The new
Sol and discovery regressionworkflow runs the real-loader tests against the pinned affected Hermes source, not a skipped optional module. Existing CI/Windows, learning, ContextEngine, dashboard, IR and packaging checks will run before merging. Final statuses will be recorded rather than inferred.Release scope
Version 0.11.2 metadata, changelog, focused notes and wiki updates are prepared. Existing release tags, Terminator artwork, schematics and the #20 live-compression implementation remain untouched. The Rust companion is version-aligned only. No live backend restart or installation has yet occurred in this repair attempt.