fix(sdk): resolve 10 pyright type errors blocking CI

[Gradata]

Summary

• CI has been red on main for multiple sessions due to pyright (basic mode) errors in SDK — unrelated to recent cloud PRs.
• Fixes all 10 reported errors; 0 errors, 7 pre-existing warnings (missing optional imports), 2070 tests pass locally.
• Unblocks PR feat: team + operator + seed + health endpoints #28 so it can merge clean.

Fixes

• RuleCache typed dict[str, list] but stores str (brain.py caches format_rules_for_prompt output). Changed signatures to str.
• Lesson._contradiction_streak set dynamically in self_improvement/rule_evolution but absent on dataclass. Added as field.
• sorted(counts, key=counts.get, ...) in behavioral_extractor fails overload resolution — wrapped with lambda.
• rule_engine.py example_draft/example_corrected — extracted to locals after None guard so pyright narrows Optional.

Test plan

• pyright src/ — 0 errors locally (was 10)
• pytest tests/ — 2070 passed, 23 skipped
• CI green on this PR

Generated with Gradata

[greptile-apps]

Gradata has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.

[coderabbitai]

📝 Walkthrough

• Fixed 10 pyright (basic mode) type errors blocking SDK CI — now 0 errors; 7 pre-existing pyright warnings remain.
• RuleCache type change (breaking): internal _cache now dict[str, str]; RuleCache.get() returns str | None; RuleCache.put(..., rules: str) — aligns cache types with format_rules_for_prompt string outputs.
• Lesson dataclass: added new field _contradiction_streak: int = 0 to avoid dynamic attribute usage in self-improvement/rule_evolution.
• behavioral_extractor: resolved overload by using key=lambda k: counts[k] when sorting counts.
• rule_engine: moved example_draft/example_corrected assignments behind None-guards so Optional types are narrowed before use.
• _embed.py: use zip(a, b, strict=False) in _cosine_distance to handle differing vector lengths safely.
• Miscellaneous non-functional/type-only tweaks: updated type imports (collections.abc.Callable), removed unused import in meta_rules, simplified control flow in rule_graph and rule_tree, and ruff lint fixes.
• Tests: local pytest run — 2,070 passed, 23 skipped.

Walkthrough

Added a _contradiction_streak field to Lesson; changed RuleCache internal storage type from list to str; minor refactors and type-annotation tweaks across rule engine, embedding, rule graph/tree, behavioral extractor, enhancements, and pubsub pipeline files; removed one unused import.

Changes

Cohort / File(s)	Summary
Datatypes src/gradata/_types.py	Added Lesson._contradiction_streak: int = 0.
Caching src/gradata/rules/cache.py	Changed _cache type dict[str, list] → dict[str, str]; get now returns `str
Rule formatting / engine src/gradata/rules/rule_engine.py	Refactored few‑shot example inclusion to use getattr locals and combined condition check; no behavior change.
Rule graph / tree src/gradata/rules/rule_graph.py, src/gradata/rules/rule_tree.py	Simplified datetime import/usage and Boolean condition expressions; removed unused TYPE_CHECKING scaffold; renamed unused loop variable to _path.
Embedding math src/gradata/_embed.py	_cosine_distance uses zip(..., strict=False) to allow differing-length vectors (truncates to shorter).
Behavioral extractor src/gradata/enhancements/behavioral_extractor.py	Changed sorting key to lambda k: counts[k] and updated forward-ref type hints to direct types.
Enhancements & utils src/gradata/enhancements/meta_rules.py, src/gradata/enhancements/self_improvement.py, src/gradata/enhancements/pubsub_pipeline.py	Removed an unused import in meta_rules; converted forward-ref annotations to direct types; switched Callable import to collections.abc.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• feat: Sim 9 engine hardening — rule scoping, convergence gate, efficiency metric #15: Modifies Lesson dataclass (adds domain_scores) — touches same class as this PR.
• feat: meta-rule discovery pipeline + behavioral extraction #19: Prior edits inside src/gradata/enhancements/behavioral_extractor.py — overlaps with sorting/type-annotation changes here.

Suggested labels

refactor

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 53.33% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely summarizes the main change: resolving 10 pyright type errors that were blocking CI, which aligns with the changeset's purpose and the detailed fixes documented.
Description check	✅ Passed	The description is directly related to the changeset, explaining the purpose of the fixes, detailing the specific issues resolved, and providing test results that validate the changes.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[cloudflare-workers-and-pages]

Deploying gradata-dashboard with    Cloudflare Pages

Latest commit:	468aaa5
Status:	🚫  Build failed.

View logs

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

src/gradata/enhancements/behavioral_extractor.py (1)

663-670: ⚠️ Potential issue | 🔴 Critical

Fix CI blocker: unresolved Ruff UP037 on quoted annotations.

Line [663] and Line [670] still use quoted Lesson annotations despite from __future__ import annotations; Ruff flags this and CI fails.

Proposed fix

 def detect_recurring_patterns(
-    lessons: list["Lesson"],
+    lessons: list[Lesson],
     min_corrections: int = 3,
 ) -> list[RecurringPattern]:
@@
-    by_cat: dict[str, list["Lesson"]] = {}
+    by_cat: dict[str, list[Lesson]] = {}

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/gradata/enhancements/behavioral_extractor.py` around lines 663 - 670, The
type annotations use quoted "Lesson" in the function signature and the by_cat
variable despite having from __future__ import annotations; remove the string
quotes so they are real names (change lessons: list["Lesson"] to lessons:
list[Lesson] and by_cat: dict[str, list["Lesson"]] to by_cat: dict[str,
list[Lesson]]), leaving other types like RecurringPattern unchanged; update any
other nearby quoted Lesson uses in that function to the unquoted form so Ruff
UP037 is resolved.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Outside diff comments:
In `@src/gradata/enhancements/behavioral_extractor.py`:
- Around line 663-670: The type annotations use quoted "Lesson" in the function
signature and the by_cat variable despite having from __future__ import
annotations; remove the string quotes so they are real names (change lessons:
list["Lesson"] to lessons: list[Lesson] and by_cat: dict[str, list["Lesson"]] to
by_cat: dict[str, list[Lesson]]), leaving other types like RecurringPattern
unchanged; update any other nearby quoted Lesson uses in that function to the
unquoted form so Ruff UP037 is resolved.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 10e43bc2-66ca-4edc-ae82-153a426631ef

📥 Commits

Reviewing files that changed from the base of the PR and between 229e7d9 and b506481.

📒 Files selected for processing (4)

• src/gradata/_types.py
• src/gradata/enhancements/behavioral_extractor.py
• src/gradata/rules/cache.py
• src/gradata/rules/rule_engine.py

📜 Review details

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (8)

• GitHub Check: Test (Python 3.11)
• GitHub Check: Test (Python 3.12)
• GitHub Check: Test (Python 3.13)
• GitHub Check: test (3.11)
• GitHub Check: test (3.12)
• GitHub Check: test (3.13)
• GitHub Check: Test (Python 3.13)
• GitHub Check: Cloudflare Pages

🧰 Additional context used

📓 Path-based instructions (1)

src/gradata/**/*.py

⚙️ CodeRabbit configuration file

src/gradata/**/*.py: This is the core SDK. Check for: type safety (from future import annotations required), no print()
statements (use logging), all functions accepting BrainContext where DB access occurs, no hardcoded paths. Severity
scoring must clamp to [0,1]. Confidence values must be in [0.0, 1.0].

Files:

• src/gradata/enhancements/behavioral_extractor.py
• src/gradata/_types.py
• src/gradata/rules/rule_engine.py
• src/gradata/rules/cache.py

🪛 GitHub Actions: SDK CI

src/gradata/enhancements/behavioral_extractor.py

[error] 663-663: ruff check reported UP037: Remove quotes from type annotation lessons: list["Lesson"],.

---

[error] 670-670: ruff check reported UP037: Remove quotes from type annotation by_cat: dict[str, list["Lesson"]] = {}.

🔇 Additional comments (4)

src/gradata/enhancements/behavioral_extractor.py (1)

650-650: Nice pyright-safe key function update.

The lambda key keeps ordering deterministic and avoids the overload issue cleanly.

src/gradata/_types.py (1)

177-177: Good fix: formalizing contradiction streak on the dataclass.

Adding _contradiction_streak to Lesson eliminates dynamic-attribute fragility and aligns with current self-improvement usage paths.

src/gradata/rules/rule_engine.py (1)

867-870: Optional narrowing is handled correctly here.

Moving attribute reads into locals and guarding before slicing fixes the type issue cleanly.

src/gradata/rules/cache.py (1)

12-29: Cache value contract alignment looks correct.

The str-typed cache storage/get/put signatures are consistent with current rule formatting output and caller expectations.

[greptile-apps]

Gradata has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (2)

src/gradata/rules/rule_graph.py (1)

232-245: ⚠️ Potential issue | 🟠 Major

Enforce confidence bounds before persistence.

confidence is written directly to SQLite without constraining/validating range. Add a clamp (or raise) to guarantee valid values on Line 244 before insert.

🔧 Proposed fix

 def store_relationship(
     db_path: str | Path,
     rule_a_id: str,
     rule_b_id: str,
     rel_type: RuleRelationType,
     confidence: float = 0.5,
 ) -> None:
     """Store a typed relationship in SQLite."""
+    confidence = max(0.0, min(1.0, confidence))
     conn = sqlite3.connect(str(db_path))
     conn.execute(
         "INSERT INTO rule_relationships "
         "(rule_a_id, rule_b_id, relationship, confidence, detected_at) "
         "VALUES (?, ?, ?, ?, ?)",

As per coding guidelines, "Confidence values must be in [0.0, 1.0]".

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/gradata/rules/rule_graph.py` around lines 232 - 245, The confidence
parameter is written to SQLite without validation; before the INSERT into
rule_relationships (the block using rel_type.value and
datetime.now(UTC).isoformat()), clamp confidence to the [0.0, 1.0] range (e.g.,
set confidence = max(0.0, min(1.0, confidence))) or alternatively raise a
ValueError if out of bounds, and then use the validated/clamped confidence in
the VALUES tuple.

src/gradata/enhancements/pubsub_pipeline.py (1)

26-26: 🧹 Nitpick | 🔵 Trivial

Consider adding type parameters for improved type precision.

The Callable type could be more specific, e.g., Callable[[dict[str, Any]], Any] to indicate the handler signature. Similarly, list[dict] at lines 34 and 48 could be list[dict[str, Any]].

While pyright passes without these (indicating they're not critical), adding them would improve type safety and IDE support.

♻️ Optional type precision improvements

-    def __init__(self):
-        self._subscribers: dict[str, list[Callable]] = {}
+    def __init__(self):
+        self._subscribers: dict[str, list[Callable[[dict[str, Any]], Any]]] = {}
         self._event_log: list[dict] = []
 
-    def subscribe(self, event_type: str, handler: Callable):
+    def subscribe(self, event_type: str, handler: Callable[[dict[str, Any]], Any]) -> None:
         if event_type not in self._subscribers:
             self._subscribers[event_type] = []
         self._subscribers[event_type].append(handler)
 
-    def emit(self, event_type: str, data: dict[str, Any] | None = None) -> list[dict]:
+    def emit(self, event_type: str, data: dict[str, Any] | None = None) -> list[dict[str, Any]]:
         """Emit an event. Returns list of stage results."""

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@src/gradata/enhancements/pubsub_pipeline.py` at line 26, Update the vague
typing to be more precise: change the declaration of self._subscribers to use a
Callable that accepts an event dict (e.g., Callable[[dict[str, Any]], Any]) and
update any occurrences of list[dict] (the buffer/queue types referenced in this
module) to list[dict[str, Any]]; ensure you import Any from typing and adjust
the type annotations on the related methods that append/consume these lists so
signatures match the new Callable[[dict[str, Any]], Any] handler shape.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Outside diff comments:
In `@src/gradata/enhancements/pubsub_pipeline.py`:
- Line 26: Update the vague typing to be more precise: change the declaration of
self._subscribers to use a Callable that accepts an event dict (e.g.,
Callable[[dict[str, Any]], Any]) and update any occurrences of list[dict] (the
buffer/queue types referenced in this module) to list[dict[str, Any]]; ensure
you import Any from typing and adjust the type annotations on the related
methods that append/consume these lists so signatures match the new
Callable[[dict[str, Any]], Any] handler shape.

In `@src/gradata/rules/rule_graph.py`:
- Around line 232-245: The confidence parameter is written to SQLite without
validation; before the INSERT into rule_relationships (the block using
rel_type.value and datetime.now(UTC).isoformat()), clamp confidence to the [0.0,
1.0] range (e.g., set confidence = max(0.0, min(1.0, confidence))) or
alternatively raise a ValueError if out of bounds, and then use the
validated/clamped confidence in the VALUES tuple.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: e8f9d89f-2e7a-4b21-9f1d-2bd45a09b5eb

📥 Commits

Reviewing files that changed from the base of the PR and between b506481 and 468aaa5.

📒 Files selected for processing (7)

• src/gradata/_embed.py
• src/gradata/enhancements/behavioral_extractor.py
• src/gradata/enhancements/meta_rules.py
• src/gradata/enhancements/pubsub_pipeline.py
• src/gradata/enhancements/self_improvement.py
• src/gradata/rules/rule_graph.py
• src/gradata/rules/rule_tree.py

💤 Files with no reviewable changes (1)

• src/gradata/enhancements/meta_rules.py

📜 Review details

⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (8)

• GitHub Check: test (3.12)
• GitHub Check: test (3.11)
• GitHub Check: test (3.13)
• GitHub Check: Python 3.12
• GitHub Check: Python 3.11
• GitHub Check: Python 3.12
• GitHub Check: Python 3.11
• GitHub Check: Cloudflare Pages

🧰 Additional context used

📓 Path-based instructions (1)

src/gradata/**/*.py

⚙️ CodeRabbit configuration file

src/gradata/**/*.py: This is the core SDK. Check for: type safety (from future import annotations required), no print()
statements (use logging), all functions accepting BrainContext where DB access occurs, no hardcoded paths. Severity
scoring must clamp to [0,1]. Confidence values must be in [0.0, 1.0].

Files:

• src/gradata/_embed.py
• src/gradata/enhancements/self_improvement.py
• src/gradata/enhancements/pubsub_pipeline.py
• src/gradata/rules/rule_tree.py
• src/gradata/enhancements/behavioral_extractor.py
• src/gradata/rules/rule_graph.py

🔇 Additional comments (10)

src/gradata/rules/rule_tree.py (2)

317-317: Unused loop variable rename is correct.

Using _path makes the intent explicit and avoids false-positive lint/type noise without changing behavior.

---

336-339: Conditional collapse preserves logic and improves readability.

The short-circuit and check is safe here and keeps contraction counting behavior intact.

src/gradata/enhancements/self_improvement.py (1)

507-507: Type annotation update is correct and pyright-friendly.

Using lesson: Lesson here is consistent with from __future__ import annotations and keeps runtime behavior unchanged while improving type checking.

src/gradata/enhancements/behavioral_extractor.py (2)

650-650: Good fix for sorted() key typing.

key=lambda k: counts[k] preserves behavior and resolves the key-function typing ambiguity cleanly.

---

663-663: Direct Lesson generics are a solid type-safety improvement.

These annotations are consistent with the module’s typing style and improve pyright inference without changing logic.

Also applies to: 670-670

src/gradata/rules/rule_graph.py (3)

17-17: Good modernization of timezone handling.

Using UTC import directly keeps timestamp code clearer and type-checker friendly.

---

212-213: Contradiction guard simplification looks correct.

The flattened conditional preserves prior semantics and improves readability.

---

216-217: Reinforcement condition refactor is clean and equivalent.

Combining category-match and overlap threshold into one guard is a solid simplification.

src/gradata/_embed.py (1)

222-222: LGTM: explicit strict=False is clear and non-breaking.

This keeps the previous truncation behavior explicit and avoids lint noise around implicit zip strictness.

src/gradata/enhancements/pubsub_pipeline.py (1)

11-12: LGTM! Import modernization follows best practices.

Moving Callable from typing to collections.abc aligns with PEP 585 and modern Python type hinting conventions (Python 3.9+). This is the recommended approach and correctly resolves type checking issues.