feat(e2e): agent playbook and real-service bootstrap states - #7
Draft
DumbMachine wants to merge 43 commits into
Draft
feat(e2e): agent playbook and real-service bootstrap states#7DumbMachine wants to merge 43 commits into
DumbMachine wants to merge 43 commits into
Conversation
Introduce an e2e/ contract for agents validating Access against live integrations (fabricate + scripts/bootstrap.sh) via the public API, with declarative workspace states and a Playwright UI screenshot harness. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Ship a curated Cloudflare API v4 preset (zones/settings/dnssec/ssl) instead of forcing fabricate mocks through rest_api, and wire e2e/states/cloudflare-zones against a local zone-audit fab profile. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Split e2e into base → fab/bootstrap → catalog layers, compose scenarios that mix them, feature playbooks for resources/chat/workflows, and a pass/fail verify harness so agents can map regressions to real data. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Address thermo-nuclear review: drop kind taxonomy in compose, derive HTTP vendors from presets, single Python e2e client, consume compose playbooks/ui_routes, and add execute_fail multi-instance checks. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Enforce declared path capabilities on workflow and agent MCP sessions so undeclared roots fail closed, while operator chat stays unscoped. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Wire capability-intersected child agent invocation from codeflows with structured-output validation, fail-closed in-flight recovery, and MCP file-root scoping so undeclared paths cannot bypass the broker.
…mple Make codeflows the primary workflow path: retire NL WORKFLOW.md demos/docs/bootstrap, deprecate legacy workflow HTTP APIs, add files.mkdir for create-if-missing durable paths, and ship a competitor-research codeflow that exercises durable files under /teams/marketing.
After merging stack-six, adapt the OpenCode agent factory to BuildScopedWorkflowConfigContent and PathCapability, and drop the unused File-Roots type leftover from the divergent MCP API.
Enable webfetch plus writable scoped MCP on codeflow child agents, always mount /mcp/workflow without blanket ReadOnly, extend run timeout for agent work, and rewrite competitor-research to gather public feedback via market-researcher into durable captures and a cited brief.
Add a read-only runtimequery adapter over resource runs, codeflows, and chat metadata, and register runtime_runs/runtime_run/runtime_failures/ runtime_chat so platform agents can diagnose and repair workflows. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Add bounded runtime_chat_list/runtime_chat summaries, agents_* and skills_* MCP tools, and a live OpenCode remediation demo harness so the platform agent can diagnose bad chats and apply prompt/skill repairs. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Replace the compact attach control's Plus icon with Paperclip so the chat page no longer crashes with Plus is not defined. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Extend the composer typeahead with scoped @ categories, / skills, and @workflow mentions that agents can inspect via runtime_runs tools.
Match the workflow detail header: breadcrumb + description stack, labeled sm actions, and no recessed panel bar around the title row. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
Open md-stream line-height so wrapped assistant text breathes, shrink streamdown inline-code so chips no longer stretch lines, and baseline-align composer mention chips with adjacent glyphs.
fix(app): align agent editor header with workflow chrome
Drop the embedded packs engine, sre-starter content, HTTP routes, and sre-demo scenario. Agents and skills come from the Agents UI or codeflow deploys instead.
Map files_read/write/patch and related durable MCP tools to clickable file cards that open the right-hand panel, add files_mkdir with parent ensure, and provision agent file roots before chat.
Let the chat header sit flush with the thread instead of a separate sidebar-colored bar.
Generalize the file dock to mixed file/run tabs, reuse RunDetail as readonly, and render runtime_* tool cards with clickable codeflow ids.
Drop the leading icon, Deployed chip, and entry-path subtitle so the name column stays quieter.
Resolve Makefile by keeping stack's demo target cleanup and our e2e compose/playbook make targets. Co-authored-by: Ratin Kumar <ratin.kumar.2k@gmail.com>
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
2 participants
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.
Summary
Layered e2e compose/playbooks/verify for agent validation against real fabricate services, plus a first-class Cloudflare OpenAPI vendor.
Also synced with
stack-1-6(files side panel, codeflow agent ops, chat chrome, etc.).Follow-up refactor addresses a thermo-nuclear code quality review:
config:+$fab.*binds; no kindeliftaxonomy incompose.pyRegisterHTTPKind+ credential codecs;DecodeConfigusesIsHTTPKindinstead of a growing case listcommon.sh; playbooks/status are Python (thin.shwrappers)playbooks/run.pyand capture loadplaybooks:/ui_routesviacompose.py show --jsoninstall_packcompose:aliases; layertype:requiredexecute_fail+ multi-instance resource playbookTest plan
origin/stack-1-6(Makefile conflict: keep stack cleanup + e2e targets)make allgo testcatalog/httpresource/kinds/opencode/files/codeflows/httpapi./e2e/bootstrap/up.sh payments-recon(+ playbooks including chat stream)./e2e/bootstrap/up.sh cloudflare-zones