Temporal + MAF agents POC: Phase 1 baseline + Phase 2 live Azure OpenAI & GitHub - #7
Open
roy2392 wants to merge 17 commits into
Open
Temporal + MAF agents POC: Phase 1 baseline + Phase 2 live Azure OpenAI & GitHub#7roy2392 wants to merge 17 commits into
roy2392 wants to merge 17 commits into
Conversation
Approved brainstorming design: planner/github/approval go live via Azure OpenAI (MAF) with real GitHub branch+plan-file+PR writes; AKS stays mock. LLM-plans/activity-executes, response_format schemas, env-selected Azure auth, PAT + owner-guard for GitHub, mocked-client tests + opt-in live smoke. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U8d6DuACgKfMFiFbaQ2AFQ
9 TDD tasks: config, MAF live seam, idempotent GitHub client, planner/ github/approval live wiring, AKS-stays-mock regression, Docker/k8s, docs + opt-in live smoke. Workflows/contracts/KEDA untouched. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U8d6DuACgKfMFiFbaQ2AFQ
The Phase-1 implementation (parent + 4 child workflows, mock agents, k8s + KEDA manifests, docker-compose, tests) was authored but never committed. Track it as the baseline before Phase-2 live wiring lands on top. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U8d6DuACgKfMFiFbaQ2AFQ
Remove agent-framework from [project].dependencies so mock-mode installs and runs without it. It remains in the [project.optional-dependencies] live extra. Also adds AZURE_OPENAI_API_VERSION to the monkeypatch.delenv loop in test_phase2_defaults to guard against a pre-set empty env var breaking the non-empty default assertion. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U8d6DuACgKfMFiFbaQ2AFQ
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U8d6DuACgKfMFiFbaQ2AFQ
…sed import RateLimitExceededException (PyGithub's 403/429 rate-limit type) is a subclass of GithubException. Catching it explicitly before the generic GithubException handler ensures rate-limit errors propagate as transient so Temporal retries them, rather than being wrapped as PermanentGitHubError and aborting the workflow. Also removes the unused STATUS_SUCCESS import in test_aks_stays_mock.py. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01U8d6DuACgKfMFiFbaQ2AFQ
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
1 participant
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.
What
Adds a self-contained
temporal-maf-agents-poc/subproject: a production-style POC where Temporal is the durable orchestration authority and Microsoft Agent Framework (MAF) is the agent runtime (used only inside Temporal activities), with one AKS worker Deployment per task queue and KEDA scaling from task-queue backlog.This branch lands two layers:
Phase 1 — topology baseline (mock agents, no credentials)
AgentOrchestratorWorkflow→ 4 child workflows (planner → github → aks → approval), one per task queue.Phase 2 — live Azure OpenAI + GitHub (this effort)
response_formatstructured output); AKS stays mock.shared/github.py), fail-closed behind an owner allowlist guard.DefaultAzureCredential/ workload identity); GitHub PAT.AGENT_MODE=mockstays the default; live wiring is opt-in. Determinism boundary preserved — no LLM/Azure/GitHub/MAF imports in workflow code.Design spec and implementation plan:
temporal-maf-agents-poc/docs/superpowers/.How it was built
Phase 2 was implemented via spec → plan → subagent-driven TDD (9 tasks), each with a per-task spec+quality review, plus a final whole-branch review. One review finding fixed: GitHub 403/429 rate-limiting is now treated as transient (was failing fast).
Tests
PYTHONPATH=src pytest→ 44 passed, 1 skipped (the skip is the creds-gated live smoke test). Includes the full time-skipping integration test in mock mode. Live paths are creds-gated and exercised only by the opt-in smoke test.🤖 Generated with Claude Code