feat(context): bind a connection profile to a project#2251
Open
PaulChen79 wants to merge 21 commits intomainfrom
Open
feat(context): bind a connection profile to a project#2251PaulChen79 wants to merge 21 commits intomainfrom
PaulChen79 wants to merge 21 commits intomainfrom
Conversation
3-tier resolver for project-aware profile lookup: read the project's `profile:` field from wren_project.yml first, fall back to the global active profile when unset. Raises SystemExit when the project pins a profile that no longer exists, so users get an actionable error instead of silent fallback. No call sites yet — pure addition. Wires up in the next commit.
Both commands now consult the project's `profile:` field (via the new resolve_profile_for_project helper) before falling back to the global active profile. Behavior is unchanged for projects without the field. The new _resolve_engine_profile helper handles both --mdl-derived and cwd-discovered project paths, so users get consistent project-aware resolution regardless of how they invoke the command.
New subcommand binds a connection profile to a project: writes `profile: <name>` plus `data_source: <profile.datasource>` into wren_project.yml. Both first-bind and rebind always overwrite the data_source from the profile (single source of truth at bind time); mismatched MDL is surfaced later by validate / build. Adds save_project_config() to wren.context for stable, ordered yaml writeback. Comments are dropped (we use PyYAML, not ruamel) — the project file's only comments are init-template copy and aren't worth a new dep.
validate now reports two new conditions: - pinned profile that no longer exists → warning (and error in --strict) - no profile pin at all → friendly info pointing to set-profile The no-pin info only fires when validation otherwise passes so error output stays uncluttered for projects with real problems.
…le step Step 2 now stops at workspace + .env (no project scaffolding yet); Step 3 creates the profile (unchanged); new Step 3.5 scaffolds the project; new Step 3.6 binds the profile via the new set-profile command, deriving data_source from the validated profile so users never edit wren_project.yml by hand. Bumps skill version 2.0 -> 2.1.
Adds a new Step 5 walking through `wren context set-profile`, explaining why explicit binding matters (multi-project safety) and when re-binding requires MDL regeneration. Renumbers downstream steps (Generate MDL, Index memory) to fit.
installation.md described the wren-onboarding skill as "project scaffolding, profile setup" — pre-change ordering. Updated both the skill summary and narrative to reflect the new flow: .env → profile → project → bind. quickstart.md Step 5 now includes the `wren context set-profile` binding step so users following the tutorial leave with a fully bound project, not one that silently falls back to global active.
github-actions
Bot
added
documentation
Contributor
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
WalkthroughThis PR adds project-scoped profile binding: new resolver honors ChangesProject Profile Binding
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Suggested reviewers
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
CI's skills/check-versions.sh enforces parity between SKILL.md frontmatter, versions.json, and index.json. The earlier commit only bumped SKILL.md; sync the registries and refresh the index.json description to match the new profile-first flow.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
core/wren/src/wren/cli.py (1)
466-468:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winUpdate datasource-missing error text to reflect project-pinned resolution.
At Line 467, the message says “active profile,” but this branch can also be using a project-pinned profile. The wording is now misleading during troubleshooting.
Suggested fix
- typer.echo("Error: no datasource in active profile.", err=True) + typer.echo( + "Error: no datasource in resolved profile (project-pinned or active).", + err=True, + )🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@core/wren/src/wren/cli.py` around lines 466 - 468, The error message in the prof_ds is None branch (the block using prof_ds, typer.echo, and raise typer.Exit) is misleading because it only mentions "active profile" even when a project-pinned profile may be in use; update the typer.echo string to reference the current profile more accurately (e.g., "Error: no datasource in the current profile (active or project-pinned)." or similar) so the message reflects both active and project-pinned profile contexts while leaving the rest of the branch (raise typer.Exit(1)) unchanged.
🧹 Nitpick comments (1)
core/wren/tests/unit/test_context_cli.py (1)
431-461: ⚡ Quick winConsider adding a test for a profile that exists but has no
datasourcefieldThe error path at
context_cli.pylines 595–601 (profile found,datasourcekey absent) has no test coverage. This is the third validation gate inset_profileand the only one not exercised.✅ Suggested test
def test_set_profile_errors_when_profile_has_no_datasource(tmp_path, monkeypatch): """Profile exists but has no datasource field → exit non-zero, message printed.""" import wren.profile as profile_mod # noqa: PLC0415 _isolate_profiles(tmp_path / "wren-home", monkeypatch) # Add a profile that is deliberately missing the datasource key profile_mod.add_profile("nodatasource", {}) proj = tmp_path / "myproj" runner.invoke(app, ["context", "init", "--empty", "--path", str(proj)]) result = runner.invoke( app, ["context", "set-profile", "nodatasource", "--path", str(proj)] ) assert result.exit_code != 0 assert "datasource" in result.output🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@core/wren/tests/unit/test_context_cli.py` around lines 431 - 461, Add a new unit test named test_set_profile_errors_when_profile_has_no_datasource in core/wren/tests/unit/test_context_cli.py that creates an isolated profile store via _isolate_profiles and uses profile_mod.add_profile("nodatasource", {}) to register a profile missing the datasource key, initializes an empty project (runner.invoke(app, ["context", "init", "--empty", "--path", str(proj)])), then calls the CLI set-profile command for "nodatasource" and asserts the command exits non‑zero and that the output contains the word "datasource" to exercise the branch in set_profile (context_cli.py) that handles profiles without a datasource.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@core/wren/src/wren/context_cli.py`:
- Around line 310-317: The call to list_profiles() in the validate logic can
raise (YAML/IO) and is currently unguarded; wrap the call to
wren.profile.list_profiles() in a try/except around the block that checks
profile_pin so exceptions are caught, and on failure append a user-friendly
semantic warning to sem_warnings (or skip the profile existence check) instead
of letting the exception propagate; reference the validate code path where
profile_pin is checked and the list_profiles() call so you locate and surround
it with try/except and include the caught exception message in the sem_warnings
entry.
In `@core/wren/src/wren/profile.py`:
- Around line 234-244: The YAML parse error is currently swallowed (except
yaml.YAMLError sets config = {}) causing silent fallback to
get_active_profile(); change this to fail-fast by re-raising or throwing a
descriptive exception when yaml.safe_load(project_yml.read_text()) raises
yaml.YAMLError. In practice, replace the except block around yaml.safe_load with
code that raises a RuntimeError or a custom exception that includes the
project_yml name/content and the original error, so the caller (and any use of
pinned_name) cannot silently proceed on malformed wren_project.yml.
---
Outside diff comments:
In `@core/wren/src/wren/cli.py`:
- Around line 466-468: The error message in the prof_ds is None branch (the
block using prof_ds, typer.echo, and raise typer.Exit) is misleading because it
only mentions "active profile" even when a project-pinned profile may be in use;
update the typer.echo string to reference the current profile more accurately
(e.g., "Error: no datasource in the current profile (active or project-pinned)."
or similar) so the message reflects both active and project-pinned profile
contexts while leaving the rest of the branch (raise typer.Exit(1)) unchanged.
---
Nitpick comments:
In `@core/wren/tests/unit/test_context_cli.py`:
- Around line 431-461: Add a new unit test named
test_set_profile_errors_when_profile_has_no_datasource in
core/wren/tests/unit/test_context_cli.py that creates an isolated profile store
via _isolate_profiles and uses profile_mod.add_profile("nodatasource", {}) to
register a profile missing the datasource key, initializes an empty project
(runner.invoke(app, ["context", "init", "--empty", "--path", str(proj)])), then
calls the CLI set-profile command for "nodatasource" and asserts the command
exits non‑zero and that the output contains the word "datasource" to exercise
the branch in set_profile (context_cli.py) that handles profiles without a
datasource.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 8415e045-6777-4314-8a7a-89893f39307a
📒 Files selected for processing (11)
core/wren/src/wren/cli.pycore/wren/src/wren/context.pycore/wren/src/wren/context_cli.pycore/wren/src/wren/profile.pycore/wren/tests/test_profile.pycore/wren/tests/unit/test_cli_profile_resolve.pycore/wren/tests/unit/test_context_cli.pydocs/core/get_started/connect.mddocs/core/get_started/installation.mddocs/core/get_started/quickstart.mdskills/wren-onboarding/SKILL.md
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@skills/index.json`:
- Line 11: The "description" field in skills/index.json references the stale
docs/get_started/connect.md path; update that string to
docs/core/get_started/connect.md so the onboarding description points to the
correct file (edit the "description" value in skills/index.json to replace
docs/get_started/connect.md with docs/core/get_started/connect.md).
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 7f33715b-0640-4e5c-814a-6c5ad7135275
📒 Files selected for processing (2)
skills/index.jsonskills/versions.json
✅ Files skipped from review due to trivial changes (1)
- skills/versions.json
resolve_profile_for_project() was silently treating a YAML parse failure as missing config and falling back to the global active profile, defeating the whole point of an explicit project-pin. Raise SystemExit with the offending path and parser message instead so users can fix the typo deterministically. Addresses CodeRabbit finding on profile.py:240.
…rors When ~/.wren/profiles.yml is unreadable (permissions, malformed YAML that escaped _load_raw, etc.), list_profiles() can raise and crash validate with a Python traceback. Wrap the lookup so we surface the failure as a warning and let the rest of validate continue running. Addresses CodeRabbit finding on context_cli.py:317.
Since dry-plan / dry-run now consult the project pin first, the 'no datasource in active profile' message is misleading when a pinned profile is missing the datasource field. Update copy to 'resolved profile (project-pinned or active)'. Addresses CodeRabbit finding on cli.py:467.
The onboarding skill description pointed at docs/get_started/connect.md but the actual file lives under docs/core/get_started/connect.md. Addresses CodeRabbit finding on index.json:11.
…field The third validation gate in set_profile (profile exists but has no datasource) had no test. Code already handles it correctly; this test locks the behavior. Addresses CodeRabbit nitpick on test_context_cli.py:431-461.
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@core/wren/src/wren/context_cli.py`:
- Around line 594-618: The set_profile flow calls list_profiles() and
save_project_config(project_path, config) without protection; wrap the
list_profiles() call in a try/except mirroring validate (catch yaml.YAMLError
and IOError/OSError) and on failure emit a user-friendly typer.echo error and
raise typer.Exit(1); likewise wrap save_project_config(...) in a try/except
catching OSError/IOError (and any write-related exceptions), emit a clear
typer.echo error message and raise typer.Exit(1) so malformed
~/.wren/profiles.yml and disk/write errors produce clean CLI errors instead of
raw tracebacks.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 817735af-8f32-40c3-8bdf-e2e622d96058
📒 Files selected for processing (6)
core/wren/src/wren/cli.pycore/wren/src/wren/context_cli.pycore/wren/src/wren/profile.pycore/wren/tests/test_profile.pycore/wren/tests/unit/test_context_cli.pyskills/index.json
🚧 Files skipped from review as they are similar to previous changes (5)
- core/wren/src/wren/profile.py
- skills/index.json
- core/wren/tests/test_profile.py
- core/wren/src/wren/cli.py
- core/wren/tests/unit/test_context_cli.py
set_profile mirrored the exact pattern validate had — bare list_profiles() and an unguarded save_project_config() — so a malformed profiles.yml or a write failure (disk full, permissions) would leak as a Python traceback. Wrap both in try/except and surface as clean CLI errors. Addresses CodeRabbit finding on context_cli.py:594-618.
Contributor
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@core/wren/tests/unit/test_context_cli.py`:
- Around line 647-658: The test currently asserts no OSError was raised and that
user-visible output exists but doesn't lock the CLI exit contract; update the
assertion for the result produced by runner.invoke(app, ["context", "validate",
"--path", str(tmp_path)]) to assert a deterministic non-zero exit code (e.g.,
assert result.exit_code == 1) so changes to error handling will fail the test;
modify the existing test near the result variable and the list_profiles
binding-check expectation to include this explicit exit-code assertion.
- Around line 371-372: The test performs multiple setup invocations using
runner.invoke(app, ["context", "init", ...]) and similar init/rebind calls
without asserting their results; change each unasserted runner.invoke call
(e.g., the initial runner.invoke(app, ["context", "init", "--empty", "--path",
str(proj)]) and other context init/rebind calls) to capture the result into a
variable and assert the invocation succeeded (for example assert
result.exit_code == 0 and/or assert expected output in result.output) so setup
failures are surfaced immediately instead of masking later test failures; update
all occurrences noted in the review (the initial init and the first rebind plus
the other listed invocations) to follow this pattern.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: a952ad1d-70ff-48c0-bfec-ed2136cb5ec9
📒 Files selected for processing (2)
core/wren/src/wren/context_cli.pycore/wren/tests/unit/test_context_cli.py
🚧 Files skipped from review as they are similar to previous changes (1)
- core/wren/src/wren/context_cli.py
- Add _invoke_ok() helper that asserts exit_code 0; route the 9 context-init / setup-rebind invocations through it so a failed scaffold step surfaces immediately instead of masking the real assertion failure further down. - Lock the exit-code contract for test_validate_handles_list_profiles_exception: warning-only path stays exit 0; --strict makes the same probe failure exit 1. Pinning both ends prevents either direction silently flipping. Addresses CodeRabbit findings on test_context_cli.py:372 and :658.
goldmedal
reviewed
May 8, 2026
Collaborator
goldmedal
left a comment
goldmedal
left a comment
There was a problem hiding this comment.
Solid PR overall — the project-pin design closes a real footgun in multi-project setups, and CodeRabbit's earlier findings are all addressed in commits 9–15. CI is green across lint/unit/memory/mysql/postgres/skill-version-parity. A few notes before merge, one of which I'd consider a follow-up worth pursuing.
1. _resolve_engine_profile couples --mdl parsing with project detection (medium)
core/wren/src/wren/cli.py:148-161: the resolver uses --mdl's path as the primary project-detection signal and only walks cwd when --mdl is absent. This re-introduces the same class of footgun the PR is trying to close.
Current behavior matrix:
--mdl |
cwd in project A | Result |
|---|---|---|
/tmp/external.json (not in any project) |
yes | global active |
<base64-string> |
yes | global active |
/other-proj/target/mdl.json |
yes | other-proj pin |
| (omitted) | yes | A's pin |
The first two rows are exactly the kind of "wrong DB silently chosen" problem this PR is designed to eliminate — just triggered by --mdl shape instead of by a stale global active.
Suggested rewrite — decouple, then fall back only when no project context exists at all:
def _resolve_engine_profile(mdl: str | None) -> tuple[str | None, dict]:
project_path = _discover_project_for_engine(mdl)
if project_path is None:
return get_active_profile()
return resolve_profile_for_project(project_path)
def _discover_project_for_engine(mdl: str | None) -> Path | None:
# 1. If --mdl is a real file, walk up looking for wren_project.yml.
# Don't hard-code parent.parent — `target/mdl.json` is the build
# default, not a contract. Users may run `--mdl <project>/build/mdl.json`.
if mdl is not None:
mdl_path = Path(mdl).expanduser()
if mdl_path.exists() and mdl_path.is_file():
for parent in mdl_path.resolve().parents:
if (parent / "wren_project.yml").exists():
return parent
if parent == Path.home() or parent == parent.parent:
break
# 2. Always try cwd discovery, regardless of whether --mdl was given.
try:
return discover_project_path()
except SystemExit:
return NoneTwo changes in one:
- Walk up from
--mdlinstead of hard-codingparent.parent— the<project>/target/mdl.jsonlayout is a build default, not a contract. - Always fall through to cwd discovery when
--mdldoesn't carry project context.--mdlbecomes a pure "which MDL artifact to load" override; project (and therefore profile) is determined independently. Active profile is reserved for "no project context anywhere" — its actual fallback role.
This also enables a coherent use case the current code blocks: running wren dry-plan --mdl /tmp/experiment.json from inside project A to test an external MDL against A's connection, without wren profile switch first.
If you'd rather keep the current "self-contained --mdl" semantics, please at least print a stderr line when falling back to active despite cwd being in a project — silent bypass is the worst of both worlds.
2. No bind-time hint when re-binding overwrites data_source and MDL is already built (low/medium)
core/wren/src/wren/context_cli.py:619-630: when old_ds != new_ds, the success line shows data_source: postgres -> duckdb, but there's no warning that an existing target/mdl.json was built for the old dialect. docs/core/get_started/connect.md does call this out, but the user only sees the doc if they read it; the CLI is silent.
Cheap UX win: when old_ds != new_ds AND (project_path / "target" / "mdl.json").exists(), append:
⚠ MDL was built for <old_ds>. Run `wren context build` to regenerate before querying.
3. set_profile hardcodes the project filename (nit)
core/wren/src/wren/context_cli.py:524 / :589: the literal "wren_project.yml" is duplicated instead of using wren.context._PROJECT_FILE. Either expose the constant (drop the underscore) or accept the duplication — current state is mildly inconsistent with the rest of context.py.
4. save_project_config round-trip not tested with custom fields (nit)
core/wren/src/wren/context.py:432-449: the field-reorder logic correctly appends unknown keys at the end, but test_set_profile_preserves_other_fields only verifies the standard fields (name, catalog, schema, schema_version). One round-trip with a custom field (e.g. tags: [foo, bar]) would lock the "extra keys are preserved" contract — cheap insurance against a future re-shuffle of _PROJECT_FIELD_ORDER accidentally dropping fields.
5. No-pin info hint suppressed when warnings exist (intentional, but worth confirming)
core/wren/src/wren/context_cli.py:344-359: the "no profile bound" hint only fires when validation has zero errors AND zero warnings. The comment says this is to keep error output uncluttered — fine if the goal is "don't pile noise on warnings," but it means a project with description warnings + no profile pin never sees the hint. Tests don't catch this because they assert on clean projects.
Worth a deliberate call: should the hint be unconditional once the structural check passes? Or always after the warning summary, on its own line so it's clearly scoped? Whichever you pick, please add a one-line comment so the next reader doesn't think it's an oversight.
Items #1 and #2 I'd address before merge; #3–#5 are fine as follow-ups.
The previous resolver only consulted cwd when --mdl was absent, so `--mdl <base64>` or `--mdl /external.json` from inside a pinned project silently bypassed the pin and fell back to global active — exactly the silent-mismatch class this PR set out to close, just triggered by --mdl shape instead of by a stale active. - Extract _discover_project_for_engine() so --mdl becomes a pure MDL artifact override; project context is determined by walking up from the MDL path AND from cwd, in that order. - Walk up directory tree from --mdl rather than hard-coding parent.parent — <project>/target/mdl.json is a build default, not a contract. - Active profile fires only when neither discovery finds a project. Addresses goldmedal review #1.
When re-binding to a profile whose datasource differs from the existing one AND target/mdl.json is already built, the manifest was emitted for the previous dialect and queries will fail until the user rebuilds. The doc mentioned this; the CLI was silent. Now the success summary includes a one-line warning pointing at `wren context build`. Addresses goldmedal review #2.
Drop the leading underscore on context.PROJECT_FILE so context_cli (in the same package) can reference it without the underscore-private signal. Replace 3 hardcoded "wren_project.yml" literals — init, set-profile presence check, and set-profile error message — with the constant. Comment / help-text mentions stay as plain strings. Addresses goldmedal review #3.
save_project_config's _PROJECT_FIELD_ORDER appends unknown keys at the end after the known fields. Existing test_set_profile_preserves_other_fields only checked the standard keys (name, catalog, schema, schema_version); this new case round-trips a custom `tags` list and `owner` string so a future re-shuffle of the field order can't silently drop user metadata. Addresses goldmedal review #4.
…an projects Previously the no-pin hint only fired when validation had zero errors AND zero warnings — so anyone working through description warnings never saw the binding nudge. Move the hint out of the success block into a separate post-warning section, gated only on no hard errors. Placed before the exit raise so --strict (which converts warnings to exit 1) still surfaces the hint. Adds two tests: hint visible under warning state, hint suppressed under hard error. Addresses goldmedal review #5.
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
wren context set-profile <name>command writes bothprofile:anddata_source:intowren_project.yml, locking a project to a specific connection profiledry-plan/dry-runnow consult the project's pinned profile before falling back to the global active one — multi-project setups stop accidentally hitting the wrong DBwren context validatehints when no profile is bound and warns when the bound profile has been removedWhy
Today, the CLI's only profile mechanism is the global
active:field in~/.wren/profiles.yml. That's fine for one-project-per-machine, but it breaks the moment you have two Wren projects pointing at different databases —wren profile switchbecomes a footgun and the SDK already supportsprofile:inwren_project.ymlwhile the CLI ignores it. This PR closes that gap with a single new command (set-profile) and a small resolver helper, so the project↔profile mapping is now an explicit first-class fact in the project file.The SDK's existing 3-tier resolution (kwarg → project pin → active) is already aligned with the new CLI behavior — no SDK changes needed.
Design decisions (locked during planning)
set-profileis new.init,profile add, andprofile switchare unchanged.set-profilesemantics: always overwritesdata_sourcefrom the bound profile (single source of truth at bind time). Mismatched MDL is surfaced later byvalidate/buildrather than blocking at bind.mkdir → fill .env → profile add → context init → set-profile → MDL gen— profile-first so the new step slots in cleanly after profile validation.validatehint tone: info (not warning) when no profile bound, so error output stays uncluttered for projects with real problems.Commit breakdown
feat(profile): add resolve_profile_for_project helperrefactor(cli): honor project-pinned profile in dry-plan and dry-runfeat(context): add wren context set-profile commandsave_project_config()helperfeat(context): hint about set-profile in validate when no profile bounddocs(skill): reorder wren-onboarding to profile-first + add set-profile stepdocs: document project↔profile binding in connect.mddocs: align installation and quickstart with profile-first onboardingTest plan
pytest core/wren/tests/— 364 passed, 1 pre-existing failure (test_validate_strict_warns) unrelated to this PRruff checkclean on every changed source/test fileWREN_HOME(set-profile binds, validate goes silent, dry-plan honors the pin)Out of scope (deliberate)
init --profile Xflag — onboarding flow now creates the profile before init runs, so the flag has no use caseprofile addcwd detection —set-profilealready covers post-add binding without polluting the global commandprofile switchwarning — keeps the global command pureprofile:correctlyWrenToolkit.from_project()— separate SDK improvement, tracked separately🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
wren context set-profileto bind a connection profile to a project; CLI now prefers a project-bound profile for project-aware commands.Bug Fixes / Improvements
dry-plan) use project-bound profiles when present.Documentation
Tests