cesnimda/jobtrackingapp
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

feat(cv-import): add source-grounded review and explicit corrections

Browse Source
This commit is contained in:
cesnimda
2026-10-26 15:56:54 +01:00
parent 9e1726c61d
commit e941b3ccb7
24 changed files with 943 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/summarizer/cv_import_ai.py
+17 -1
View File
@@ -37,6 +37,21 @@ otherSections.N.title or otherSections.N.items.N. Relationships across ambiguous
FIELD = re.compile(r"^(?:contact\.(?:fullName|headline|email|phone|location|website|linkedIn|gitHub)|contact\.links\.\d{1,3}\.(?:label|url)|(?:summary|skills|interests|awards|publications|organisations|references)\.\d{1,3}|(?:jobs|education|certifications|projects|languages)\.\d{1,3}\.(?:title|company|name|institution|qualification|qualificationLevel|issuer|role|location|start|end|date|level|notes|(?:bullets|skills|details)\.\d{1,3})|otherSections\.\d{1,3}\.(?:title|items\.\d{1,3}))$")
def _supported_field(field):
parts = field.split(".")
fields = {
"jobs": ({"title", "company", "location", "start", "end"}, {"bullets", "skills"}),
"education": ({"institution", "qualification", "qualificationLevel", "location", "start", "end"}, {"details"}),
"certifications": ({"name", "issuer", "location", "date"}, {"details"}),
"projects": ({"name", "role", "location", "start", "end"}, {"bullets", "skills"}),
"languages": ({"name", "level", "notes"}, set()),
}
if parts[0] not in fields:
return bool(FIELD.fullmatch(field))
scalar, lists = fields[parts[0]]
return bool(FIELD.fullmatch(field)) and ((len(parts) == 3 and parts[2] in scalar) or (len(parts) == 4 and parts[2] in lists))
def _validate_config(config):
url = urlsplit(config.base_url)
if url.scheme != "http" or url.username or url.password or url.query or url.fragment or url.path not in ("", "/") or not config.model or len(config.model) > 200:
@@ -88,7 +103,8 @@ def analyse(blocks, config=None, generate=None):
for item in parsed["suggestions"]:
if not isinstance(item, dict) or set(item) != {"field", "value", "block_id", "quote"} or any(not isinstance(value, str) or not value.strip() or len(value) > 8000 for value in item.values()):
raise ValueError("Invalid model suggestion.")
if not FIELD.fullmatch(item["field"]) or item["field"] in seen or item["block_id"] not in by_id or clean(item["quote"]) not in clean(by_id[item["block_id"]]) or clean(item["value"]) not in clean(item["quote"]):
grounded_value = re.search(r"(?<![\w+#.])" + re.escape(clean(item["value"])) + r"(?![\w+#.])", clean(item["quote"]))
if not _supported_field(item["field"]) or item["field"] in seen or item["block_id"] not in by_id or clean(item["quote"]) not in clean(by_id[item["block_id"]]) or not grounded_value:
raise ValueError("Unsupported model suggestion.")
seen.add(item["field"])
suggestions.append({**item, "method": "model-assisted", "review_state": "needs_review"})