fix: ensure cache files are UTF-8 encoded

fanoutqa/eval/llm.py

@@ -42,7 +42,7 @@ async def get_llm_factuality(question: DevQuestion, answer: str, cache_key=None)
         ans_hash = hashlib.sha256(answer.encode()).hexdigest()[:8]
         cache_filename = LLM_CACHE_DIR / f"factual-{cache_key}-{question.id}-{ans_hash}.txt"
         if cache_filename.exists():
-            return cache_filename.read_text()
+            return cache_filename.read_text(encoding="utf-8")
 
     # ask the LLM if it is subjective
     prompt = factuality_prompt(question.question, str_answer(question.answer), answer)
@@ -51,5 +51,5 @@ async def get_llm_factuality(question: DevQuestion, answer: str, cache_key=None)
 
     if cache_key:
         # noinspection PyUnboundLocalVariable
-        cache_filename.write_text(resp)
+        cache_filename.write_text(resp, encoding="utf-8")
     return resp

fanoutqa/wiki.py

@@ -76,7 +76,7 @@ def wiki_content(doc: Evidence) -> str:
     # get the cached content, if available
     cache_filename = WIKI_CACHE_DIR / f"{doc.pageid}-dated.md"
     if cache_filename.exists():
-        return cache_filename.read_text()
+        return cache_filename.read_text(encoding="utf-8")
 
     # otherwise retrieve it from Wikipedia
     resp = wikipedia.get("", params={"format": "json", "action": "parse", "oldid": doc.revid, "prop": "text"})
@@ -90,5 +90,5 @@ def wiki_content(doc: Evidence) -> str:
 
     # MD it, cache it, and return
     text = markdownify(html)
-    cache_filename.write_text(text)
+    cache_filename.write_text(text, encoding="utf-8")
     return text

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "fanoutqa"
-version = "1.1.0"
+version = "1.1.1"
 authors = [
     { name = "Andrew Zhu", email = "[email protected]" },
     { name = "Alyssa Hwang", email = "[email protected]" },