feat(codegen): add unified multi-target CLI architecture

[pyramation]

feat(codegen): add unified multi-target CLI architecture

Summary

Adds a unified multi-target CLI generator that produces a single CLI for projects with multiple GraphQL API endpoints (e.g., auth.localhost, members.localhost, app.localhost). Instead of generating separate CLIs per target, this generates one CLI with target-namespaced commands (auth:user list, members:member create, app:car get).

Key pieces:

• resolveBuiltinNames() — collision detection that auto-renames builtin commands when they clash with target names (e.g., auth → credentials, context → env), with user-configurable overrides via builtinNames config
• Multi-target executor — routes to the correct ORM client per target, resolves per-target endpoints from appstash context
• Multi-target command map — target:command namespacing for all schema-derived commands; builtin commands (context, auth/credentials) remain top-level
• Multi-target context command — context create prompts for per-target endpoints with defaults baked from config
• --save-token flag — auth target mutations can auto-save tokens to credentials store
• generateMulti() wiring — new unifiedCli option that skips per-target CLI generation, collects pipeline data, and generates a single unified CLI
• Multi-target docs generation — README, AGENTS.md, MCP tools, and skills files generated for unified CLI with target-namespaced examples and collision-aware builtin command names

This PR also includes the prior work from PRs #725 (CLI codegen) and #728 (docs generation) which this branch builds on.

Updates since last revision

• Renamed infraNames → builtinNames across the entire codebase (BuiltinNames interface, resolveBuiltinNames() function, CliConfig.builtinNames property, all generators, all tests). This is a purely mechanical rename for API clarity — "builtin" better communicates that these are the built-in commands that ship with every generated CLI, as opposed to the schema-derived ones.

• Added multi-target documentation generators that produce docs reflecting the unified CLI's target-namespaced command structure:

  • generateMultiTargetReadme() — README.md showing target:command syntax, per-target endpoint setup via context create, and collision-renamed builtin commands
  • generateMultiTargetAgentsDocs() — AGENTS.md structured for LLM consumption with per-target tool definitions and multi-target workflows
  • getMultiTargetCliMcpTools() — MCP tool definitions for all targets with target-prefixed tool names
  • generateMultiTargetSkills() — per-target skill files with target-prefixed file names

  These are wired into generateMulti() in generate.ts — when useUnifiedCli is enabled and docsConfig is present, docs are written alongside the generated CLI code. 5 new snapshot tests added (284 tests total, 118 snapshots).

Review & Testing Checklist for Human

• Verify generated code actually compiles: The snapshot tests show generated TypeScript, but it has never been executed. Try running the codegen against a real multi-target schema and tsc the output. Check imports resolve correctly (especially ../../executor paths from commands/auth/user.ts).
• Test collision detection edge cases: What happens if a user sets builtinNames: { auth: 'credentials' } but also has a target named credentials? The current code doesn't detect secondary collisions. Also verify the default renaming works (auth → credentials, context → env).
• Verify skipCli / runOrm logic change: In generate.ts, the condition for runOrm changed from runReactQuery || runCli || ... to runReactQuery || !!config.cli || .... This means ORM is now generated even when skipCli: true is passed (which is correct for multi-target, but verify it doesn't break single-target behavior).
• Check appstash dependency: The generated executor calls store.getTargetEndpoint(targetName) which was added in dev-utils PR Move req object and ApiModule to server/src/types #64. Verify that PR is merged and published (appstash@0.5.0 or similar) before merging this.
• Test isAuthTarget detection: The code hardcodes isAuthTarget = name === 'auth' in generateMulti(). This means only targets literally named auth get --save-token on mutations. Consider if this should be configurable or detected differently.
• Verify multi-target docs accuracy: The docs generators read docsConfig from the first target (configs[names[0]]?.docs), assuming all targets share the same docs config. Test with a real multi-target project to ensure docs correctly show target-namespaced commands, collision-renamed builtin commands, and per-target endpoint setup instructions.

Notes

• Devin run: https://app.devin.ai/sessions/e9c668834d394cdc8ebcb371b6ebf430
• Requested by: @pyramation
• Dependencies: Requires appstash@0.5.0+ (dev-utils PR Move req object and ApiModule to server/src/types #64) for getTargetEndpoint() / setTargetEndpoint() API
• Test coverage: 8 new snapshot tests for multi-target CLI scenarios, 6 unit tests for resolveBuiltinNames(), 5 new snapshot tests for multi-target docs. All 284 tests pass (118 snapshots).
• Breaking changes: None for single-target usage. Multi-target is opt-in via unifiedCli option in generateMulti().

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring