LLM attribute calculation

mustache.py

@@ -0,0 +1,12 @@
+from typing import Dict, Any
+import chevron
+
+
+def prepare_and_render_mustache(template_text: str, record_dict: Dict[str, Any]) -> str:
+    comb_dict = {
+        "rd": record_dict,
+    }
+    return chevron.render(
+        template=template_text,
+        data=comb_dict,
+    )

requirements.txt

@@ -9,9 +9,7 @@ aiohappyeyeballs==2.4.3
     #   -r requirements/exec-env-requirements.txt
     #   aiohttp
 aiohttp==3.11.7
-    # via
-    #   -r requirements/exec-env-requirements.txt
-    #   openai
+    # via -r requirements/exec-env-requirements.txt
 aiosignal==1.3.1
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -23,6 +21,7 @@ anyio==4.6.2.post1
     #   -r requirements/exec-env-requirements.txt
     #   groq
     #   httpx
+    #   openai
 async-timeout==5.0.1
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -68,6 +67,8 @@ charset-normalizer==3.4.0
     # via
     #   -r requirements/exec-env-requirements.txt
     #   requests
+chevron==0.14.0
+    # via -r requirements/requirements.in
 click==8.1.7
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -99,6 +100,7 @@ distro==1.9.0
     # via
     #   -r requirements/exec-env-requirements.txt
     #   groq
+    #   openai
 docopt==0.6.2
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -144,6 +146,7 @@ httpx==0.27.2
     # via
     #   -r requirements/exec-env-requirements.txt
     #   groq
+    #   openai
 idna==3.10
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -167,6 +170,8 @@ jinja2==3.1.4
     # via
     #   -r requirements/exec-env-requirements.txt
     #   spacy
+jiter==0.8.2
+    # via openai
 joblib==1.4.2
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -254,8 +259,8 @@ numpy==1.23.4
     #   spacy
     #   textacy
     #   thinc
-openai==0.27.7
-    # via -r requirements/exec-env-requirements.txt
+openai==1.59.7
+    # via -r requirements/requirements.in
 packaging==24.2
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -297,6 +302,7 @@ pydantic==2.7.4
     #   -r requirements/requirements.in
     #   confection
     #   groq
+    #   openai
     #   spacy
     #   thinc
     #   weasel
@@ -343,7 +349,6 @@ requests==2.31.0
     #   -r requirements/exec-env-requirements.txt
     #   google-search-results
     #   lexmo
-    #   openai
     #   spacy
     #   sumy
     #   textacy
@@ -380,6 +385,7 @@ sniffio==1.3.1
     #   anyio
     #   groq
     #   httpx
+    #   openai
 soupsieve==2.6
     # via
     #   -r requirements/exec-env-requirements.txt
@@ -466,6 +472,7 @@ typing-extensions==4.12.2
     #   cloudpathlib
     #   groq
     #   multidict
+    #   openai
     #   pydantic
     #   pydantic-core
     #   sqlalchemy

requirements/exec-env-requirements.txt

@@ -171,7 +171,7 @@ numpy==1.23.4
     #   spacy
     #   textacy
     #   thinc
-openai==0.27.7
+# openai==0.27.7
     # via -r exec-env-requirements.in
 packaging==24.2
     # via

requirements/requirements.in

@@ -1,3 +1,5 @@
 -r exec-env-requirements.txt
 pydantic==2.7.4
-spacy[ja]==3.7.5
+spacy[ja]==3.7.5
+chevron==0.14.0
+openai==1.59.7

run_ac.py

@@ -2,6 +2,7 @@
 import requests
 import spacy
 import sys
+from mustache import prepare_and_render_mustache
 from spacy.tokens import DocBin
 
 
@@ -18,6 +19,8 @@ def get_check_data_type_function(data_type):
         return [str], __check_data_type_text
     elif data_type == "EMBEDDING_LIST":
         return [list], __check_data_type_embedding_list
+    elif data_type == "LLM_RESPONSE":
+        return [str], __check_data_type_text
     else:
         raise ValueError(f"Unknown data type: {data_type}")
 
@@ -105,7 +108,7 @@ def parse_data_to_record_dict(record_chunk):
 
     # This import statement will always be highlighted as a potential error, as during devtime,
     # the script `labeling_functions` does not exist. It will be inserted at runtime
-    from attribute_calculators import ac
+    import attribute_calculators
 
     vocab = spacy.blank(iso2_code).vocab
 
@@ -123,11 +126,15 @@ def parse_data_to_record_dict(record_chunk):
     amount = len(record_dict_list)
     __print_progress(0.0)
     for record_dict in record_dict_list:
+        attribute_calculators.USER_PROMPT = prepare_and_render_mustache(
+            attribute_calculators.USER_PROMPT, record_dict
+        )
+
         idx += 1
         if idx % progress_size == 0:
             progress = round(idx / amount, 2)
             __print_progress(progress)
-        attr_value = ac(record_dict["data"])
+        attr_value = attribute_calculators.ac(record_dict["data"])
         if not check_data_type(attr_value):
             raise ValueError(
                 f"Attribute value `{attr_value}` is of type {type(attr_value)}, "