Global chat job code

[hanna-paasivirta]

Short Description

Testing improvements and bug fixes for a set of user scenarios targeting job code edits relating to just the step that the user is viewing (i.e. not several steps, not another step that the user is not viewing, and not the workflow structure).

Fixes #490

Implementation Details

New acceptance specs for global_chat

• Added one_shot_workflows/ specs exercising the planner path on from-scratch workflow requests of varying ambiguity.
• Added job_code/ specs (8 scenarios) exercising the job-code path on existing workflows — first-turn questions, multi-turn conversations, log-driven debugging, "what does this step do", and code edits with varying history depth.

Architecture / planning docs

• Added agent-team-architecture-plan/ with the design write-ups for the four test tiers and an example spec format, so future contributors know where unit / service / integration / acceptance tests should live.

Bug fix in planner job-code stitching

• Fixed services/global_chat/planner.py: when the planner's job_key needed fuzzy-matching to find the YAML job, the original key was still being passed to stitch_job_code (which requires exact match), so the stitch silently failed and code
  edits were lost. Now captures the matched key and uses it for stitching. Affected both single and parallel job-code-tool execution paths.

Prompt fixes

• Prompt tweak in global_chat/prompts.yaml to firm up router step selection

AI Usage

Please disclose how you've used AI in this work (it's cool, we just want to know!):

• Code generation (copilot but not intellisense)
• Learning or fact checking
• Strategy / design
• Optimisation / refactoring
• Translation / spellchecking / doc gen
• Other
• I have not used AI

You can read more details in our Responsible AI Policy