Repository navigation
Platform: remove Tavall AI runtime/provider ownership from Function Catalog - #13
tjXJNOOBIE wants to merge 4 commits into
Conversation
|
@codex review current exact head |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 527442ab5f
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".
| - `agent-runtime` | ||
| - `codex-agent-provider` | ||
|
|
||
| are removed from the active Gradle build and source tree. They were execution-runtime/provider implementations rather than function-catalog infrastructure. |
There was a problem hiding this comment.
When readers consult the repository documentation after this removal, docs/TAVALL_AGENT_RUNTIME_ARCHITECTURE.md still describes agent-runtime as an active Function Catalog module and documents deleted APIs such as AIAgentRuntime and AIAgentProvider. A repository-wide symbol search confirms that guide now contains the only surviving references, so it contradicts the new ownership boundary and directs implementers toward nonexistent types; remove it or replace its contents with a pointer to the Tavall AI-owned architecture.
Useful? React with 👍 / 👎.
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 527442ab5f
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".
|
|
||
| This removal is paired with Tavall AI PR #10, **Runtime: move model execution and Codex provider into Tavall AI**. | ||
|
|
||
| The Function Catalog removal PR must remain Draft / blocked from staging promotion until the Tavall AI replacement passes its owning Java 25 local verification and DEVELOPMENT Codex runtime acceptance. Source removal in this repository is not evidence that the replacement runtime is operational. |
There was a problem hiding this comment.
This prerequisite is only documented, not enforced. I checked .github/workflows/gradle.yml, .github/workflows/publish.yml, and .github/TAVALL_STAGING.md: the required check validates only this repository, publishing can be dispatched after the change reaches main, and staging promotion is merely marked manual. Consequently, this exact head can pass every repository gate and be promoted with both runtime/provider implementations removed even if Tavall AI #10 has not passed Java 25 verification or DEVELOPMENT Codex acceptance. Add a required cross-repository acceptance/status gate, or another mechanically enforced promotion condition, rather than relying solely on Draft state.
Useful? React with 👍 / 👎.
1 participant
Summary
Remove actual AI/model runtime and Codex provider implementation from Function Catalog now that the replacement ownership boundary exists in Tavall AI PR #10.
This is a focused child of Function Catalog staging PR #10 (
staging/platform). It remains Draft and must not promote ahead of the Tavall AI replacement validation.Removed Function Catalog modules
agent-runtimecodex-agent-providerThey are removed from:
settings.gradle.kts;Function Catalog does not retain compatibility copies. Two live runtime/provider implementations would recreate the ownership ambiguity this migration is fixing.
Function Catalog boundary after this PR
Function Catalog remains authoritative for:
AIFunctionCatalog/ scopedAIFunctionCatalogView;Tavall AI consumes Function Catalog through
org.tavall:ai-core.Replacement in Tavall AI
Tavall AI PR #10 (
Runtime: move model execution and Codex provider into Tavall AI) adds:tavall-ai-runtime-model-executionTavallAgent+ provider execution definition;tavall-ai-runtime-codexMoving the adapter does not move ambient process authority: Tavall Cloud/runtime host remains authoritative for DEVELOPMENT eligibility, workspace leases, process isolation/supervision, executables/credentials, resource authority, cancellation, and fencing.
Sibling integration work
repository_staging_*functions + GitHub provider + automatic MCP projection.All three children edit the staging future tree. Reconciliation into
staging/platformmust combine their intendedsettings.gradle.kts/build.gradle.ktschanges rather than allowing one sibling to overwrite another.Validation truth
No Java 25 build has been executed from this ChatGPT connector.
On this branch, the equivalent current exact-head acceptance is:
After sibling #12 integrates/rebases, the canonical gate becomes:
Cross-repository removal acceptance additionally requires Tavall AI #10 to pass exact-head Java 25 local verification and DEVELOPMENT Codex runtime acceptance before this removal is considered promotion-ready.
State: IMPLEMENTED_NEEDS_VALIDATION
Owner: chatgpt-interactive
Scope: remove Function Catalog-owned Tavall AI model runtime and Codex provider
Relationships: CHILD_OF Function Catalog staging PR #10; SIBLING_OF Function Catalog #11/#12; DEPENDS_ON Tavall AI #10 replacement
Validation: SOURCE_REVIEWED; FUNCTION_CATALOG_JAVA25_LOCAL_CI_PENDING; TAVALL_AI_10_JAVA25_LOCAL_CI_PENDING; CODEX_DEVELOPMENT_E2E_PENDING
Next Action: validate Function Catalog exact head, validate Tavall AI #10 replacement, then reconcile all Function Catalog staging siblings together before promotion.