reformat

_secret.template.py

@@ -1,11 +1,7 @@
 from util import zip_api
 
-api_test = zip_api(
-    api_base="",
-    api_key="",
-    model=""
-)
+api_test = zip_api(api_base="", api_key="", model="")
 
 cmd = """
 OPENAI_API_BASE="" OPENAI_API_KEY="" CHEAP_MODEL="" STRONG_MODEL="" run.sh
-"""
+"""

app.py

@@ -9,17 +9,21 @@ from taskNonAI import extract_url, file_to_html, compile_pdf
 
 ## load data
 from _data_test import mock_jd, mock_cv
+
 ## ui
 import gradio as gr
+
 ## dependency
 from pypandoc.pandoc_download import download_pandoc
+
 ## std
 import os
 import json
 
-logger = mylogger(__name__,'%(asctime)s:%(levelname)s:%(message)s')
+logger = mylogger(__name__, "%(asctime)s:%(levelname)s:%(message)s")
 info = logger.info
 
+
 def init():
     os.system("shot-scraper install -b firefox")
     download_pandoc()
@@ -27,24 +31,32 @@ def init():
 
 ## Config Functions
 
-def set_same_cheap_strong(set_same:bool, cheap_base, cheap_key):
+
+def set_same_cheap_strong(set_same: bool, cheap_base, cheap_key):
     # setup_zone = gr.Accordion("AI setup (OpenAI-compatible LLM API)", open=True)
     if set_same:
-        return (gr.Textbox(value=cheap_base, label="API Base", interactive=False),
-                gr.Textbox(value=cheap_key, label="API key", type="password", interactive=False),
-                gr.Textbox(value=cheap_model, label="Model ID", interactive=False),
-                # setup_zone,
-                )
+        return (
+            gr.Textbox(value=cheap_base, label="API Base", interactive=False),
+            gr.Textbox(
+                value=cheap_key, label="API key", type="password", interactive=False
+            ),
+            gr.Textbox(value=cheap_model, label="Model ID", interactive=False),
+            # setup_zone,
+        )
     else:
-        return (gr.Textbox(value=cheap_base, label="API Base", interactive=True),
-                gr.Textbox(value=cheap_key, label="API key", type="password", interactive=True),
-                gr.Textbox(value=cheap_model, label="Model ID", interactive=True),
-                # setup_zone,
-                )
-    
+        return (
+            gr.Textbox(value=cheap_base, label="API Base", interactive=True),
+            gr.Textbox(
+                value=cheap_key, label="API key", type="password", interactive=True
+            ),
+            gr.Textbox(value=cheap_model, label="Model ID", interactive=True),
+            # setup_zone,
+        )
+
 
 ## Main Functions
 
+
 def prepare_input(jd_info, cv_file: str, cv_text):
     if jd_info:
         if is_valid_url(jd_info):
@@ -66,8 +78,9 @@ def prepare_input(jd_info, cv_file: str, cv_text):
         cv = mock_cv
     return jd, cv
 
+
 def run_refine(api_base, api_key, api_model, jd_info, cv_text):
-    jd,cv=jd_info,cv_text
+    jd, cv = jd_info, cv_text
     cheapAPI = {"base": api_base, "key": api_key, "model": api_model}
     taskAI = TaskAI(cheapAPI, temperature=0.2, max_tokens=2048)  # max_tokens=2048
     info("API initialized")
@@ -78,6 +91,7 @@ def run_refine(api_base, api_key, api_model, jd_info, cv_text):
     for result in gen:
         yield result
 
+
 def run_compose(api_base, api_key, api_model, min_jd, min_cv):
     strongAPI = {"base": api_base, "key": api_key, "model": api_model}
     taskAI = TaskAI(strongAPI, temperature=0.6, max_tokens=4000)
@@ -87,22 +101,29 @@ def run_compose(api_base, api_key, api_model, min_jd, min_cv):
         result += response.delta
         yield result
 
+
 def finalize_letter_txt(api_base, api_key, api_model, debug_CoT):
     cheapAPI = {"base": api_base, "key": api_key, "model": api_model}
     taskAI = TaskAI(cheapAPI, temperature=0.2, max_tokens=2048)
     info("Finalizing letter ...")
-    result=""
+    result = ""
     for response in taskAI.purify_letter(full_text=debug_CoT):
         result += response.delta
         yield result
 
+
 def finalize_letter_pdf(api_base, api_key, api_model, jd, cv, cover_letter_text):
     cheapAPI = {"base": api_base, "key": api_key, "model": api_model}
     taskAI = TaskAI(cheapAPI, temperature=0.1, max_tokens=100)
     meta_data = next(taskAI.get_jobapp_meta(JD=jd, CV=cv))
     pdf_context = json.loads(meta_data)
     pdf_context["letter_body"] = cover_letter_text
-    return meta_data, compile_pdf(pdf_context,tmpl_path="typst/template_letter.tmpl",output_path=f"/tmp/cover_letter_by_{pdf_context['applicantFullName']}_to_{pdf_context['companyFullName']}.pdf")
+    return meta_data, compile_pdf(
+        pdf_context,
+        tmpl_path="typst/template_letter.tmpl",
+        output_path=f"/tmp/cover_letter_by_{pdf_context['applicantFullName']}_to_{pdf_context['companyFullName']}.pdf",
+    )
+
 
 with gr.Blocks(
     title=DEMO_TITLE,
@@ -116,26 +137,28 @@ with gr.Blocks(
 
     with gr.Row():
         with gr.Column(scale=1):
-            with gr.Accordion("AI setup (OpenAI-compatible LLM API)", open=False) as setup_zone:
-                is_debug = gr.Checkbox( label="Debug Mode", value=IS_DEBUG)
-                
+            with gr.Accordion(
+                "AI setup (OpenAI-compatible LLM API)", open=False
+            ) as setup_zone:
+                is_debug = gr.Checkbox(label="Debug Mode", value=IS_DEBUG)
+
                 gr.Markdown(
                     "**Cheap AI**, an honest format converter and refiner, extracts essential info from job description and résumé, to reduce subsequent cost on Strong AI."
                 )
                 with gr.Group():
-                    cheap_base = gr.Textbox(
-                        value=CHEAP_API_BASE, label="API Base"
+                    cheap_base = gr.Textbox(value=CHEAP_API_BASE, label="API Base")
+                    cheap_key = gr.Textbox(
+                        value=CHEAP_API_KEY, label="API key", type="password"
                     )
-                    cheap_key = gr.Textbox(value=CHEAP_API_KEY, label="API key", type="password")
                     cheap_model = gr.Textbox(value=CHEAP_MODEL, label="Model ID")
                 gr.Markdown(
                     "---\n**Strong AI**, a thoughtful wordsmith, generates perfect cover letters to make both you and recruiters happy."
                 )
-                is_same_cheap_strong = gr.Checkbox(label="the same as Cheap AI", value=False, container=False)
+                is_same_cheap_strong = gr.Checkbox(
+                    label="the same as Cheap AI", value=False, container=False
+                )
                 with gr.Group():
-                    strong_base = gr.Textbox(
-                        value=STRONG_API_BASE, label="API Base"
-                    )
+                    strong_base = gr.Textbox(value=STRONG_API_BASE, label="API Base")
                     strong_key = gr.Textbox(
                         value=STRONG_API_KEY, label="API key", type="password"
                     )
@@ -150,7 +173,7 @@ with gr.Blocks(
                 )
             with gr.Group():
                 gr.Markdown("## Applicant - CV / Résumé")
-            # with gr.Row():
+                # with gr.Row():
                 cv_file = gr.File(
                     label="Allowed formats: " + " ".join(CV_EXT),
                     file_count="single",
@@ -168,7 +191,6 @@ with gr.Blocks(
                     min_jd = gr.TextArea(label="Reformatted Job Description")
                     min_cv = gr.TextArea(label="Reformatted CV / Résumé")
             with gr.Accordion("Expert Zone", open=False) as expert_zone:
-                
                 debug_CoT = gr.Textbox(label="Chain of Thoughts")
                 debug_jobapp = gr.Textbox(label="Job application meta data")
             cover_letter_text = gr.Textbox(label="Cover Letter")
@@ -179,27 +201,54 @@ with gr.Blocks(
                 type="filepath",
             )
             infer_btn = gr.Button("Go!", variant="primary")
-            
-    is_same_cheap_strong.change(fn= set_same_cheap_strong, 
-                                    inputs=[is_same_cheap_strong, cheap_base, cheap_key, cheap_model],
-                                    outputs=[strong_base, strong_key, strong_model])
-
-    infer_btn.click(fn= set_same_cheap_strong, 
-                    inputs=[is_same_cheap_strong, cheap_base, cheap_key, cheap_model],
-                    outputs=[strong_base, strong_key, strong_model]
+
+    is_same_cheap_strong.change(
+        fn=set_same_cheap_strong,
+        inputs=[is_same_cheap_strong, cheap_base, cheap_key, cheap_model],
+        outputs=[strong_base, strong_key, strong_model],
+    )
+
+    infer_btn.click(
+        fn=set_same_cheap_strong,
+        inputs=[is_same_cheap_strong, cheap_base, cheap_key, cheap_model],
+        outputs=[strong_base, strong_key, strong_model],
     ).success(
-        fn=prepare_input,
-        inputs=[jd_info, cv_file, cv_text],
-        outputs=[jd_info, cv_text]
+        fn=prepare_input, inputs=[jd_info, cv_file, cv_text], outputs=[jd_info, cv_text]
     ).success(
         fn=run_refine,
         inputs=[cheap_base, cheap_key, cheap_model, jd_info, cv_text],
         outputs=[min_jd, min_cv],
-    ).success(fn=lambda:[gr.Accordion("Expert Zone", open=True),gr.Accordion("Reformatting", open=False)],inputs=None, outputs=[expert_zone, reformat_zone]
-    ).success(fn=run_compose, inputs=[strong_base, strong_key, strong_model, min_jd, min_cv], outputs=[debug_CoT]                      
-    ).success(fn=lambda:gr.Accordion("Expert Zone", open=False),inputs=None, outputs=[expert_zone]
-    ).success(fn=finalize_letter_txt, inputs=[cheap_base, cheap_key, cheap_model, debug_CoT], outputs=[cover_letter_text]
-    ).success(fn=finalize_letter_pdf, inputs=[cheap_base, cheap_key, cheap_model, jd_info, cv_text, cover_letter_text], outputs=[debug_jobapp, cover_letter_pdf])
+    ).success(
+        fn=lambda: [
+            gr.Accordion("Expert Zone", open=True),
+            gr.Accordion("Reformatting", open=False),
+        ],
+        inputs=None,
+        outputs=[expert_zone, reformat_zone],
+    ).success(
+        fn=run_compose,
+        inputs=[strong_base, strong_key, strong_model, min_jd, min_cv],
+        outputs=[debug_CoT],
+    ).success(
+        fn=lambda: gr.Accordion("Expert Zone", open=False),
+        inputs=None,
+        outputs=[expert_zone],
+    ).success(
+        fn=finalize_letter_txt,
+        inputs=[cheap_base, cheap_key, cheap_model, debug_CoT],
+        outputs=[cover_letter_text],
+    ).success(
+        fn=finalize_letter_pdf,
+        inputs=[
+            cheap_base,
+            cheap_key,
+            cheap_model,
+            jd_info,
+            cv_text,
+            cover_letter_text,
+        ],
+        outputs=[debug_jobapp, cover_letter_pdf],
+    )
 
 
 if __name__ == "__main__":

taskAI.py

@@ -83,7 +83,7 @@ class TaskAI(OpenAILike):
             return window_size
 
         checkAPI(api_base, api_key)
-        
+
         super().__init__(
             api_base=api["base"],
             api_key=api["key"],

taskNonAI.py

@@ -1,5 +1,6 @@
 import pypandoc
 import typst
+
 ## stdlib
 import subprocess
 import json
@@ -40,26 +41,37 @@ def extract_url(url: str) -> Optional[str]:
             f"Please try copy-paste as input. Failed to extract content from: {url}. Didn't find content from given URL!"
         )
 
-def _date()->str:
+
+def _date() -> str:
     current_date = datetime.now()
     return current_date.strftime(
-        f"%B %d{'th' if 4 <= current_date.day <= 20 or 24 <= current_date.day <= 30 else ['st', 'nd', 'rd'][current_date.day % 10 - 1]} , %Y")
+        f"%B %d{'th' if 4 <= current_date.day <= 20 or 24 <= current_date.day <= 30 else ['st', 'nd', 'rd'][current_date.day % 10 - 1]} , %Y"
+    )
+
+
+def _typst_escape(s) -> str:
+    return str(s).replace("@", "\@").replace("#", "\#")
+
 
-def _typst_escape(s)->str:
-    return str(s).replace('@','\@').replace('#','\#')
- 
-def compile_pdf(context: dict, tmpl_path: str, output_path="/tmp/cover_letter.pdf", is_debug=False)->list[str]:
-    letter_src_filepath = 'typst/letter.typ'
-    with open(tmpl_path, "r", encoding='utf8') as f:
+def compile_pdf(
+    context: dict, tmpl_path: str, output_path="/tmp/cover_letter.pdf", is_debug=False
+) -> list[str]:
+    letter_src_filepath = "typst/letter.typ"
+    with open(tmpl_path, "r", encoding="utf8") as f:
         tmpl = Template(f.read())
     context = {k: _typst_escape(v) for k, v in context.items()}
-    context.update({'date_string': _date()})
+    context.update({"date_string": _date()})
     letter_typ = tmpl.safe_substitute(context)
-    with open(letter_src_filepath, 'w', encoding='utf8') as f:
+    with open(letter_src_filepath, "w", encoding="utf8") as f:
         f.write(letter_typ)
-    typst.compile(letter_src_filepath, output_path, root=Path('./typst/'), font_paths=[Path('./fonts/')])
+    typst.compile(
+        letter_src_filepath,
+        output_path,
+        root=Path("./typst/"),
+        font_paths=[Path("./fonts/")],
+    )
     # os.remove(letter_src_filepath)
     if is_debug:
         return [letter_src_filepath, output_path]
     else:
-        return [output_path]
+        return [output_path]

util.py

@@ -22,7 +22,7 @@ def mylogger(name, format, level=logging.INFO):
     return logger
 
 
-def count_token(text, encoding="cl100k_base")->int:
+def count_token(text, encoding="cl100k_base") -> int:
     return len(tiktoken.get_encoding(encoding).encode(text))
 
 
@@ -41,9 +41,13 @@ def is_valid_openai_api_key(api_base: str, api_key: str) -> bool:
 
     return response.status_code == 200
 
+
 def checkAPI(api_base: str, api_key: str):
     if not is_valid_openai_api_key(api_base, api_key):
-        raise ValueError("Invalid API key or less possibly OpenAI's (or AI provider's) fault. Did you setup your AI APIs properly? If you don't have any API key, try get one from https://beta.openai.com/account/api-keys")
+        raise ValueError(
+            "Invalid API key or less possibly OpenAI's (or AI provider's) fault. Did you setup your AI APIs properly? If you don't have any API key, try get one from https://beta.openai.com/account/api-keys"
+        )
+
 
 def zip_api(api_base: str, api_key: str, model: str) -> dict[str, str]:
     return {"base": api_base, "key": api_key, "model": model}