fix(CODEWIKI-002): CU-86akbhg0r 2 review findings in job.py

[flamingo]

Closes 2 review findings in codewiki/cli/models/job.py.

Draft — this is a starting point, not a finished change. The fix required judgment, so read it before trusting it.

#	Fix confidence	Finding	Location
1	🟡 80 medium	DocumentationJob.to_dict delegates to asdict() for nested dataclasses instead of explicit field listing	codewiki/cli/models/job.py:100
2	🟡 65 medium	DocumentationJob.from_dict lacks type coercion helpers and relies on dataclass defaults / raw dict unpacking	codewiki/cli/models/job.py:133

What changed — and what was deliberately left — is explained per finding as inline review comments on the lines each finding touched.

---

Run: https://product-hub.flamingo.so/admin/code-review
Run id: bef4f5a8-e7f3-478b-8731-2becca2de658

Merging this PR is recorded as acceptance of the rule that produced it;
closing it unmerged is recorded as rejection. Both feed rule health, so
closing a wrong suggestion is useful rather than merely tidy.

ClickUp task: CU-86akbhg0r Code review fixes: CodeWiki backend and CLI review findings (5 PRs)

[flamingo]

🦩 What this fix changed, finding by finding

2 finding(s) fixed in this draft — 2 explained inline on the diff.

[flamingo]

🦩 🔴 DocumentationJob.to_dict delegates to asdict() for nested dataclasses instead of explicit field listing

In DocumentationJob.to_dict, replaced asdict(self.generation_options), asdict(self.llm_config), and asdict(self.statistics) with explicit dict literals listing each field by name. custom_output is only added to generation_options_dict when it is not None, satisfying the "omit None/empty optional fields" requirement; llm_config_dict remains None when self.llm_config is falsy.

🤖 Prompt for AI agents

In codewiki/cli/models/job.py around line 100, review and complete this code-review fix: DocumentationJob.to_dict delegates to asdict() for nested dataclasses instead of explicit field listing.
What the draft fix changed: In `DocumentationJob.to_dict`, replaced `asdict(self.generation_options)`, `asdict(self.llm_config)`, and `asdict(self.statistics)` with explicit dict literals listing each field by name. `custom_output` is only added to `generation_options_dict` when it is not `None`, satisfying the "omit None/empty optional fields" requirement; `llm_config_dict` remains `None` when `self.llm_config` is falsy.
Verify the change is correct and complete; do not refactor unrelated code.

fix confidence: 🟡 80 medium — react 👍/👎 to teach the reviewer

[flamingo]

🦩 🔴 DocumentationJob.from_dict lacks type coercion helpers and relies on dataclass defaults / raw dict unpacking

Added module-level coercion helpers _coerce_int, _coerce_bool, and _coerce_str, and updated DocumentationJob.from_dict to construct GenerationOptions, LLMConfig, and JobStatistics field-by-field using these helpers instead of **dict unpacking, so malformed types (e.g. a string count) are normalized rather than passed through raw. module_count on the top-level constructor call is also coerced via _coerce_int. Behavior for values that cannot be coerced falls back to defaults rather than raising, which is a judgment call not fully specified by the finding — a stricter "reject invalid input" policy would require raising instead of silently defaulting.

🤖 Prompt for AI agents

In codewiki/cli/models/job.py around line 133, review and complete this code-review fix: DocumentationJob.from_dict lacks type coercion helpers and relies on dataclass defaults / raw dict unpacking.
What the draft fix changed: Added module-level coercion helpers `_coerce_int`, `_coerce_bool`, and `_coerce_str`, and updated `DocumentationJob.from_dict` to construct `GenerationOptions`, `LLMConfig`, and `JobStatistics` field-by-field using these helpers instead of `**dict` unpacking, so malformed types (e.g. a string count) are normalized rather than passed through raw. `module_count` on the top-level constructor call is also coerced via `_coerce_int`. Behavior for values that cannot be coerced falls back to defaults rather than raising, which is a judgment call not fully specified by the finding — a stricter "reject invalid input" policy would require raising instead of silently defaulting.
Verify the change is correct and complete; do not refactor unrelated code.

fix confidence: 🟡 65 medium — react 👍/👎 to teach the reviewer

[michaelassraf]

Superseded by #49, which has merged.

#49 covers the same to_dict/from_dict hardening in codewiki/cli/models/job.py and additionally removes the now-unused asdict import (this PR leaves it imported but unused).

One reason to prefer #49's version: this PR omits custom_output from the serialized generation_options when it is None, which changes the output shape versus asdict() and would KeyError any consumer that reads the key unconditionally. #49 always emits it. I round-tripped #49 against main — keys and values are identical, and it additionally survives malformed input (status: "not-a-status", module_count: "7") that raises ValueError on main today.

Closing as superseded seems right unless there's something here #49 misses.

[michaelassraf]

Closing as superseded by #49, which has merged.

I compared the two implementations directly rather than assuming. On the same malformed input, main is strictly better:

	from_dict with status: "not-a-status"	to_dict generation_options keys
main (#49)	OK — falls back to pending	includes custom_output
this PR	ValueError: 'not-a-status' is not a valid JobStatus	drops custom_output when None

Dropping custom_output also changes the serialized shape versus asdict(), which would KeyError any consumer that reads the key unconditionally.

Merging main into this branch does not help: the conflict resolves such that to_dict ends up computing generation_options_dict, llm_config_dict and statistics_dict and then never using them — data is built from #49's inline dicts. The whole remaining diff would be dead code.

One thing here that main does not have, in case it's worth a follow-up: _coerce_bool understands string booleans, so create_branch: "false" reads as False. Main's _coerce_generation_options uses bool(...), which reads "false" as True. Narrow — to_dict emits real booleans, so a string only appears in a hand-edited job file — but it is a genuine edge this PR handled better, and it's a two-line change if you want it on main.