Merge pull request #4 from Jwaminju/part-of-support-smolagent

agent/handler.py

@@ -13,25 +13,48 @@ from agent.workflow import (
 )
 from pr_generator.searcher import find_reference_pr_simple_stream
 from translator.content import get_full_prompt, get_content, preprocess_content
+from translator.project_config import get_available_projects, get_project_config
 
 
 # State management
 class ChatState:
     def __init__(self):
         self.step = "welcome"  # welcome -> find_files -> translate -> create_github_pr
+        
+        # Transient state (reset on restart)
+        self.selected_project = "transformers"
         self.target_language = "ko"
         self.k_files = 10
         self.files_to_translate = []
         self.additional_instruction = ""
         self.current_file_content = {"translated": ""}
-        self.pr_result = None  # Store PR creation result
-        # GitHub configuration
-        self.github_config = {
-            "token": "",
-            "owner": "",
-            "repo_name": "",
-            "reference_pr_url": "https://github.com/huggingface/transformers/pull/24968",
+        self.pr_result = None
+        
+        # Persistent settings (preserved across restarts)
+        self.persistent_settings = {
+            "anthropic_api_key": "",
+            "github_config": {
+                "token": "",
+                "owner": "",
+                "repo_name": "",
+                "reference_pr_url": "",
+            }
         }
+    
+    def reset_transient_state(self):
+        """Reset only the workflow state, keep persistent settings"""
+        self.step = "welcome"
+        self.selected_project = "transformers"
+        self.target_language = "ko"
+        self.k_files = 10
+        self.files_to_translate = []
+        self.additional_instruction = ""
+        self.current_file_content = {"translated": ""}
+        self.pr_result = None
+    
+    @property
+    def github_config(self):
+        return self.persistent_settings["github_config"]
 
 
 state = ChatState()
@@ -53,25 +76,41 @@ def _extract_content_for_display(content: str) -> str:
 
 
 def get_welcome_message():
-    """Initial welcome message with file finding controls"""
+    """Initial welcome message with project selection"""
     return """**👋 Welcome to 🌐 Hugging Face i18n Translation Agent!**
 
 I'll help you find files that need translation and translate them in a streamlined workflow.
 
-**🔎 Let's start by finding files that need translation.**
+**🎯 First, select which project you want to translate:**
 
-Use the **`Quick Controls`** on the right or **ask me `what`, `how`, or `help`** to get started.
+Use the **`Quick Controls`** on the right to select a project, or **ask me `what`, `how`, or `help`** to get started.
 """
 
 
-def process_file_search_handler(lang: str, k: int, history: list) -> tuple:
+def process_file_search_handler(project: str, lang: str, k: int, history: list) -> tuple:
     """Process file search request and update Gradio UI components."""
     global state
+    state.selected_project = project
     state.target_language = lang
     state.k_files = k
     state.step = "find_files"
 
-    status_report, files_list = report_translation_target_files(lang, k)
+    try:
+        status_report, files_list = report_translation_target_files(project, lang, k)
+    except Exception as e:
+        if "rate limit" in str(e).lower():
+            response = f"""❌ **GitHub API Rate Limit Exceeded**
+
+{str(e)}
+
+**💡 To fix this:**
+1. Set GitHub Token in Configuration panel above
+2. Click "💾 Save Configuration" 
+3. Try "Find Files" again"""
+            history.append(["File search request", response])
+            return history, "", update_status(), gr.Tabs(selected=0), gr.update(choices=[]), gr.update(visible=False)
+        else:
+            raise  # Re-raise non-rate-limit errors
     state.files_to_translate = (
         [file[0] for file in files_list]
         if files_list
@@ -87,8 +126,10 @@ def process_file_search_handler(lang: str, k: int, history: list) -> tuple:
 """
 
     if state.files_to_translate:
+        config = get_project_config(state.selected_project)
         for i, file in enumerate(state.files_to_translate, 1):
-            response += f"\n{i}. `{file}`"
+            file_link = f"{config.repo_url}/blob/main/{file}"
+            response += f"\n{i}. [`{file}`]({file_link})"
 
         # if len(state.files_to_translate) > 5:
         #     response += f"\n... and {len(state.files_to_translate) - 5} more files"
@@ -100,14 +141,13 @@ def process_file_search_handler(lang: str, k: int, history: list) -> tuple:
     # Add to history
     history.append(["Please find files that need translation", response])
     cleared_input = ""
-    selected_tab = 1 if state.files_to_translate else 0
 
     # 드롭다운 choices로 쓸 파일 리스트 반환 추가
     return (
         history,
         cleared_input,
         update_status(),
-        gr.Tabs(selected=selected_tab),
+        gr.Tabs(),  # Don't change tab
         update_dropdown_choices(state.files_to_translate),
     )
 
@@ -116,7 +156,30 @@ def update_dropdown_choices(file_list):
     return gr.update(choices=file_list, value=None)
 
 
-def start_translation_process():
+def confirm_and_go_translate_handler(history):
+    """Confirm selection and go to translate tab"""
+    global state
+    
+    response = f"✅ **Selection confirmed!**\n\n🎯 **Project:** {state.selected_project}\n🌍 **Language:** {state.target_language}\n\n**➡️ Go to Tab 2 to start translation.**"
+    history.append(["Confirm selection", response])
+    return history, "", update_status(), gr.Tabs(selected=1)
+
+
+def confirm_translation_and_go_upload_handler(history):
+    """Confirm translation and go to upload PR tab"""
+    global state
+    
+    if not state.current_file_content.get("translated"):
+        response = "❌ No translation available. Please complete translation first."
+        history.append(["Upload PR request", response])
+        return history, "", update_status(), gr.Tabs()
+    
+    response = f"✅ **Translation confirmed!**\n\n📄 **File:** `{state.files_to_translate[0] if state.files_to_translate else 'Unknown'}`\n\n**➡️ Go to Tab 3 to upload PR.**"
+    history.append(["Upload PR request", response])
+    return history, "", update_status(), gr.Tabs(selected=2)
+
+
+def start_translation_process(force_retranslate=False):
     """Start the translation process for the first file"""
     if not state.files_to_translate:
         return "❌ No files available for translation.", ""
@@ -125,8 +188,8 @@ def start_translation_process():
 
     # Call translation function (simplified for demo)
     try:
-        translated = translate_docs_interactive(
-            state.target_language, [[current_file]], state.additional_instruction
+        status, translated = translate_docs_interactive(
+            state.target_language, [[current_file]], state.additional_instruction, state.selected_project, force_retranslate
         )
 
         state.current_file_content = {"translated": translated}
@@ -138,19 +201,22 @@ def start_translation_process():
         p.parent.mkdir(parents=True, exist_ok=True)
         p.write_text(translated, encoding="utf-8")
 
-        original_file_link = (
-            "https://github.com/huggingface/transformers/blob/main/" + current_file
-        )
+        config = get_project_config(state.selected_project)
+        original_file_link = f"{config.repo_url}/blob/main/{current_file}"
         print("Compeleted translation:\n")
         print(translated)
         print("----------------------------")
-        response = (
-            f"""🔄 Translation for: `{current_file}`\n"""
-            "**📄 Original Content Link:**\n"
-            ""
-            f"{original_file_link}\n"
-            "**🌐 Translated Content:**\n"
-        )
+        
+        # Different response format for existing vs new translation
+        if isinstance(status, str) and "Existing translation loaded" in status:
+            response = f"{status}\n**📄 Original Content Link:** {original_file_link}\n\n**🌐 Translated Content:**"
+        else:
+            response = (
+                f"""🔄 Translation for: `{current_file}`\n"""
+                f"**📄 Original Content Link:** {original_file_link}\n\n"
+                f"{status}\n\n"
+                "**🌐 Translated Content:**"
+            )
         return response, translated
 
 
@@ -188,7 +254,12 @@ Currently available actions with quick controls:
     else:
         return """I understand you want to work on translations! 
 
-To get started, please use the controls above to configure your translation settings and find files that need translation.
+**Two ways to get started:**
+
+1. **🔍 Find Files first** - Use Tab 1 to discover files that need translation
+2. **🚀 Direct Translation** - Go to Tab 2 and enter a file path directly (e.g., `docs/source/en/model_doc/bert.md`)
+
+Make sure to configure your API keys in the Configuration panel above.
 """
 
 
@@ -226,12 +297,12 @@ def handle_user_message(message, history):
 
 def update_status():
     if state.step == "welcome":
-        return """
+        return f"""
         <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 10px;padding: 10px; background: rgba(0, 0, 0, 0.25); border-radius: 8px;">
             <div><strong>🔄 Step:</strong> Welcome</div>
+            <div><strong>🎯 Project:</strong> {state.selected_project}</div>
             <div><strong>📁 Files:</strong> 0</div>
-            <div><strong>🌍 Language:</strong> ko</div>
-            <div><strong>⏳ Progress:</strong> Ready</div>
+            <div><strong>🌍 Language:</strong> {state.target_language}</div>
         </div>
         """
 
@@ -267,6 +338,7 @@ def update_status():
     status_html = f"""
     <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 10px; padding: 10px; background: rgba(0, 0, 0, 0.25); border-radius: 8px;">
         <div><strong>🔄 Step:</strong> {step_map.get(state.step, state.step)}</div>
+        <div><strong>🎯 Project:</strong> {state.selected_project}</div>
         <div><strong>📁 Files:</strong> {len(state.files_to_translate)}</div>
         <div><strong>🌍 Language:</strong> {state.target_language}</div>
         <div><strong>⏳ Progress:</strong> {progress_map.get(state.step, 'In progress')}</div>
@@ -284,26 +356,66 @@ def sync_language_displays(lang):
     return lang
 
 
-def update_github_config(token, owner, repo, reference_pr_url):
-    """Update GitHub configuration settings."""
+def update_project_selection(project, history):
+    """Update state when project is selected"""
     global state
+    state.selected_project = project
+    response = f"Selection confirmed: 🎯 Project → **{project}**"
+    history.append(["Project selection", response])
+    return history, "", update_status()
+
+
+def update_language_selection(lang, history):
+    """Update state when language is selected"""
+    global state
+    state.target_language = lang
+    response = f"Selection confirmed: 🌍 Language → **{lang}**"
+    history.append(["Language selection", response])
+    return history, "", update_status(), lang
+
+
+def update_persistent_config(anthropic_key, github_token, github_owner, github_repo, reference_pr_url, history):
+    """Update persistent configuration settings."""
+    global state
+    
+    # Update API keys
+    if anthropic_key:
+        state.persistent_settings["anthropic_api_key"] = anthropic_key
+        os.environ["ANTHROPIC_API_KEY"] = anthropic_key
+    
+    if github_token:
+        os.environ["GITHUB_TOKEN"] = github_token
+
+    # Get default reference PR URL from project config if not provided
+    if not reference_pr_url and state.selected_project:
+        try:
+            config = get_project_config(state.selected_project)
+            reference_pr_url = config.reference_pr_url
+        except:
+            pass
+
+    # Save GitHub configuration to persistent settings
+    state.persistent_settings["github_config"].update({
+        "token": github_token or "",
+        "owner": github_owner or "",
+        "repo_name": github_repo or "",
+        "reference_pr_url": reference_pr_url or "",
+    })
+
+    # Build response message based on what was configured
+    response = "✅ Configuration saved!"
+    if github_owner and github_repo:
+        response += f" GitHub: {github_owner}/{github_repo}"
+    elif anthropic_key:
+        response += " Anthropic API key updated."
+    
+    history.append(["Configuration update", response])
+    return history, "", update_status()
 
-    # Set GitHub token in environment variables
-    if token:
-        os.environ["GITHUB_TOKEN"] = token
-
-    # Save GitHub configuration to state
-    state.github_config.update(
-        {
-            "token": token,
-            "owner": owner,
-            "repo_name": repo,
-            "reference_pr_url": reference_pr_url
-            or state.github_config["reference_pr_url"],
-        }
-    )
 
-    return f"✅ GitHub configuration updated: {owner}/{repo}"
+def update_github_config(token, owner, repo, reference_pr_url):
+    """Legacy function for backward compatibility."""
+    return update_persistent_config("", token, owner, repo, reference_pr_url)
 
 
 def update_prompt_preview(language, file_path, additional_instruction):
@@ -319,7 +431,7 @@ def update_prompt_preview(language, file_path, additional_instruction):
             translation_lang = language
         
         # Get sample content (first 500 characters)
-        content = get_content(file_path)
+        content = get_content(file_path, state.selected_project)
         to_translate = preprocess_content(content)
         
         # Truncate for preview
@@ -330,7 +442,10 @@ def update_prompt_preview(language, file_path, additional_instruction):
         
         return prompt
     except Exception as e:
-        return f"Error generating prompt preview: {str(e)}"
+        error_str = str(e)
+        if "Failed to retrieve content from the URL" in error_str:
+            return f"❌ **File not found:** `{file_path}`\n\n💡 **Please check:**\n1. Is this file in the **{state.selected_project}** project?\n2. Use \"🔍 Find Files to Translate\" to see available files\n3. Verify the file path is correct"
+        return f"Error generating prompt preview: {error_str}"
 
 
 def send_message(message, history):
@@ -339,14 +454,39 @@ def send_message(message, history):
 
 
 # Button handlers with tab switching
-def start_translate_handler(history, anthropic_key, file_to_translate, additional_instruction=""):
+def start_translate_handler(history, file_to_translate, additional_instruction="", force_retranslate=False):
+    # Use persistent anthropic key
+    anthropic_key = state.persistent_settings["anthropic_api_key"]
+    if not anthropic_key:
+        response = "❌ Please set Anthropic API key in Configuration panel first."
+        history.append(["Translation request", response])
+        return history, "", update_status(), gr.Tabs(), gr.update(), gr.update()
+    
     os.environ["ANTHROPIC_API_KEY"] = anthropic_key
     
+    # Check if file path is provided
+    if not file_to_translate or not file_to_translate.strip():
+        response = "❌ Please select a file from the dropdown or enter a file path to translate."
+        history.append(["Translation request", response])
+        return history, "", update_status(), gr.Tabs(), gr.update(), gr.update()
+    
     state.additional_instruction = additional_instruction
     state.files_to_translate = [file_to_translate]
-    new_hist, cleared_input = handle_user_message("start translation", history)
-    selected_tabs = 2 if state.current_file_content["translated"] else 0
-    return new_hist, cleared_input, update_status(), gr.Tabs(selected=selected_tabs)
+    state.step = "translate"
+    
+    # Start translation directly
+    if force_retranslate:
+        history.append(["Translation request", "🔄 **Force retranslation started...**"])
+    response, translated = start_translation_process(force_retranslate)
+    history.append(["", response])
+    if translated:
+        history.append(["", translated])
+    
+    # Update button text and show confirm button after translation
+    start_btn_text = "🔄 Retranslation" if state.current_file_content["translated"] else "🚀 Start Translation"
+    confirm_btn_visible = bool(state.current_file_content["translated"])
+    
+    return history, "", update_status(), gr.Tabs(), gr.update(value=start_btn_text), gr.update(visible=confirm_btn_visible)
 
 
 def approve_handler(history, owner, repo, reference_pr_url):
@@ -354,18 +494,34 @@ def approve_handler(history, owner, repo, reference_pr_url):
     global state
     state.step = "create_github_pr"
 
-    # Update github config from the latest UI values
-    state.github_config["owner"] = owner
-    state.github_config["repo_name"] = repo
-    state.github_config["reference_pr_url"] = reference_pr_url
-
-    # Validate GitHub configuration
-    github_config = state.github_config
-    if not all([github_config.get("token"), owner, repo]):
-        response = "❌ GitHub configuration incomplete. Please provide GitHub Token, Owner, and Repository Name in Tab 3."
+    # Check all required GitHub configuration at once
+    github_config = state.persistent_settings["github_config"]
+    missing_config = []
+    
+    if not github_config.get("token"):
+        missing_config.append("GitHub Token")
+    if not owner:
+        missing_config.append("GitHub Owner")
+    if not repo:
+        missing_config.append("Repository Name")
+    
+    if missing_config:
+        config = get_project_config(state.selected_project)
+        repo_name = config.repo_url.split('/')[-1]  # Extract repo name from URL
+        response = f"❌ Please set the following in Configuration panel first: {', '.join(missing_config)}\n\n💡 **Note:** GitHub Owner/Repository should be your fork of [`{repo_name}`]({config.repo_url}) (e.g., Owner: `your-username`, Repository: `{repo_name}`)"
         history.append(["GitHub PR creation request", response])
         return history, "", update_status()
 
+    # Update reference PR URL (can be set per PR)
+    if reference_pr_url:
+        state.persistent_settings["github_config"]["reference_pr_url"] = reference_pr_url
+
+    # Use persistent settings
+    github_config = state.persistent_settings["github_config"]
+
+    # Initialize response variable
+    response = ""
+    
     # If reference PR is not provided, use the agent to find one
     if not github_config.get("reference_pr_url"):
         response = "🤖 **Reference PR URL not found. The agent will now search for a suitable one...**"
@@ -421,6 +577,7 @@ def approve_handler(history, owner, repo, reference_pr_url):
             translated_content=translated_content,
             github_config=state.github_config,
             en_title=file_name,
+            project=state.selected_project,
         )
         response += f"\n{pr_response}"
     else:
@@ -431,9 +588,23 @@ def approve_handler(history, owner, repo, reference_pr_url):
 
 
 def restart_handler(history):
-    """Resets the state and UI."""
+    """Resets the workflow state but preserves persistent settings."""
     global state
+    # Backup persistent settings
+    backup_settings = state.persistent_settings.copy()
+    
+    # Reset state
     state = ChatState()
+    
+    # Restore persistent settings
+    state.persistent_settings = backup_settings
+    
+    # Restore environment variables
+    if backup_settings["anthropic_api_key"]:
+        os.environ["ANTHROPIC_API_KEY"] = backup_settings["anthropic_api_key"]
+    if backup_settings["github_config"]["token"]:
+        os.environ["GITHUB_TOKEN"] = backup_settings["github_config"]["token"]
+    
     welcome_msg = get_welcome_message()
     new_hist = [[None, welcome_msg]]
     return new_hist, "", update_status(), gr.Tabs(selected=0)

agent/toctree_handler.py

@@ -4,9 +4,17 @@ from typing import Dict, List, Any
 import os
 
 class TocTreeHandler:
-    def __init__(self):
-        self.en_toctree_url = "https://raw.githubusercontent.com/huggingface/transformers/main/docs/source/en/_toctree.yml"
-        self.ko_toctree_url = "https://raw.githubusercontent.com/huggingface/transformers/main/docs/source/ko/_toctree.yml"
+    def __init__(self, project: str = "transformers"):
+        from translator.project_config import get_project_config
+        self.project = project
+        self.project_config = get_project_config(project)
+        
+        # Extract repository path from config
+        repo_path = self.project_config.repo_url.replace("https://github.com/", "")
+        
+        # Build project-specific URLs
+        self.en_toctree_url = f"https://raw.githubusercontent.com/{repo_path}/main/docs/source/en/_toctree.yml"
+        self.ko_toctree_url = f"https://raw.githubusercontent.com/{repo_path}/main/docs/source/ko/_toctree.yml"
         self.local_docs_path = "docs/source/ko"
     
     def fetch_toctree(self, url: str) -> Dict[str, Any]:
@@ -245,7 +253,8 @@ Korean title:"""
         translation_result: dict, 
         filepath: str, 
         pr_agent, 
-        github_config: dict
+        github_config: dict,
+        project: str = "transformers"
     ) -> dict:
         """Update toctree after successful translation PR.
         

agent/workflow.py

@@ -11,7 +11,7 @@ from translator.content import (
     llm_translate,
     preprocess_content,
 )
-from translator.retriever import report, get_github_issue_open_pr
+from translator.retriever import report, get_github_issue_open_pr, get_github_repo_files
 # GitHub PR Agent import
 try:
     from pr_generator.agent import GitHubPRAgent
@@ -26,19 +26,23 @@ from logger.github_logger import GitHubLogger
 
 
 def report_translation_target_files(
-    translate_lang: str, top_k: int = 1
+    project: str, translate_lang: str, top_k: int = 1
 ) -> tuple[str, list[list[str]]]:
     """Return the top-k files that need translation, excluding files already in progress.
 
     Args:
+        project: Project to translate (e.g., "transformers", "smolagents")
         translate_lang: Target language to translate
         top_k: Number of top-first files to return for translation. (Default 1)
     """
-    # Get files in progress
-    docs_in_progress, pr_info_list = get_github_issue_open_pr(translate_lang)
-
-    # Get all available files for translation
-    all_status_report, all_filepath_list = report(translate_lang, top_k * 2)  # Get more to account for filtering
+    # Get repo files once to avoid duplicate API calls
+    all_repo_files = get_github_repo_files(project)
+    
+    # Get all available files for translation using the file list
+    all_status_report, all_filepath_list = report(project, translate_lang, top_k * 2, all_repo_files)  # Get more to account for filtering
+    
+    # Get files in progress using the same file list
+    docs_in_progress, pr_info_list = get_github_issue_open_pr(project, translate_lang, all_repo_files)
 
     # Filter out files that are already in progress
     available_files = [f for f in all_filepath_list if f not in docs_in_progress]
@@ -52,29 +56,30 @@ def report_translation_target_files(
     if docs_in_progress:
         status_report += f"\n\n🤖 Found {len(docs_in_progress)} files in progress for translation:"
         for i, file in enumerate(docs_in_progress):
-            status_report += f"\n{i+1}. `{file}`: {pr_info_list[i]}"
+            status_report += f"\n{i+1}. [`{file}`]({pr_info_list[i]})"
         status_report += f"\n\n📋 Showing {len(filepath_list)} available files (excluding in-progress):"
 
     return status_report, [[file] for file in filepath_list]
 
 
-def translate_docs(lang: str, file_path: str, additional_instruction: str = "") -> tuple[str, str]:
+def translate_docs(lang: str, file_path: str, additional_instruction: str = "", project: str = "transformers", force_retranslate: bool = False) -> tuple[str, str]:
     """Translate documentation."""
-    # Check if translation already exists
+    # Check if translation already exists (unless force retranslate is enabled)
     translation_file_path = (
         Path(__file__).resolve().parent.parent
         / f"translation_result/{file_path}"
     )
 
-    if translation_file_path.exists():
+    if not force_retranslate and translation_file_path.exists():
         print(f"📄 Found existing translation: {translation_file_path}")
         with open(translation_file_path, "r", encoding="utf-8") as f:
             existing_content = f.read()
         if existing_content.strip():
-            return "Existing translation loaded (no tokens used). If you want to translate again, please restart the gradio app.", existing_content
+            existing_msg = f"♻️ **Existing translation loaded** (no tokens used)\n📁 **File:** `{file_path}`\n📅 **Loaded from:** `{translation_file_path}`\n💡 **To retranslate:** Check 'Force Retranslate' option."
+            return existing_msg, existing_content
 
     # step 1. Get content from file path
-    content = get_content(file_path)
+    content = get_content(file_path, project)
     to_translate = preprocess_content(content)
 
     # step 2. Prepare prompt with docs content
@@ -97,7 +102,7 @@ def translate_docs(lang: str, file_path: str, additional_instruction: str = "")
 
 
 def translate_docs_interactive(
-    translate_lang: str, selected_files: list[list[str]], additional_instruction: str = ""
+    translate_lang: str, selected_files: list[list[str]], additional_instruction: str = "", project: str = "transformers", force_retranslate: bool = False
 ) -> tuple[str, str]:
     """Interactive translation function that processes files one by one.
 
@@ -111,14 +116,22 @@ def translate_docs_interactive(
     # Start with the first file
     current_file = file_paths[0]
 
-    status = f"✅ Translation completed: `{current_file}` → `{translate_lang}`\n\n"
-    callback_result, translated_content = translate_docs(translate_lang, current_file, additional_instruction)
-    status += f"💰 Used token and cost: \n```\n{callback_result}\n```"
+    callback_result, translated_content = translate_docs(translate_lang, current_file, additional_instruction, project, force_retranslate)
+    
+    # Check if existing translation was loaded
+    if isinstance(callback_result, str) and "Existing translation loaded" in callback_result:
+        status = callback_result  # Use the existing translation message
+    else:
+        if force_retranslate:
+            status = f"🔄 **Force Retranslation completed**: `{current_file}` → `{translate_lang}`\n\n"
+        else:
+            status = f"✅ Translation completed: `{current_file}` → `{translate_lang}`\n\n"
+        status += f"💰 Used token and cost: \n```\n{callback_result}\n```"
 
     print(callback_result)
     print(status)
 
-    return translated_content
+    return status, translated_content
 
 
 def generate_github_pr(
@@ -127,6 +140,7 @@ def generate_github_pr(
     translated_content: str = None,
     github_config: dict = None,
     en_title: str = None,
+    project: str = "transformers",
 ) -> str:
     """Generate a GitHub PR for translated documentation.
 
@@ -144,7 +158,7 @@ def generate_github_pr(
         return "❌ GitHub PR Agent is not available. Please install required libraries."
 
     if not github_config:
-        return "❌ GitHub configuration not provided."
+        return "❌ GitHub configuration not provided. Please set up GitHub token, owner, and repository in Configuration panel."
 
     # Validate required configuration
     required_fields = ["token", "owner", "repo_name", "reference_pr_url"]
@@ -153,7 +167,7 @@ def generate_github_pr(
     ]
 
     if missing_fields:
-        return f"❌ Missing required configuration: {', '.join(missing_fields)}. Please provide these values."
+        return f"❌ Missing required GitHub configuration: {', '.join(missing_fields)}\n\n💡 Go to Configuration panel and set:\n" + "\n".join([f"  • {field}" for field in missing_fields])
 
     # Set token in environment for the agent.
     os.environ["GITHUB_TOKEN"] = github_config["token"]
@@ -166,29 +180,39 @@ def generate_github_pr(
                 / f"translation_result/{filepath}"
             )
             if not translation_file_path.exists():
-                return f"❌ Translation file not found: {translation_file_path}"
+                return f"❌ Translation file not found: {translation_file_path}\n\n💡 Please complete translation first in Tab 2 for file: {filepath}"
 
             with open(translation_file_path, "r", encoding="utf-8") as f:
                 translated_content = f.read()
 
         if not translated_content or not translated_content.strip():
-            return "❌ Translated content is empty."
+            return f"❌ Translated content is empty for file: {filepath}\n\n💡 Please complete translation first in Tab 2."
 
         # Execute GitHub PR Agent
+        # Get base repository from project config
+        from translator.project_config import get_project_config
+        project_config = get_project_config(project)
+        base_repo_path = project_config.repo_url.replace("https://github.com/", "")
+        base_owner, base_repo = base_repo_path.split("/")
+
         print(f"🚀 Starting GitHub PR creation...")
         print(f"   📁 File: {filepath}")
         print(f"   🌍 Language: {target_language}")
         print(f"   📊 Reference PR: {github_config['reference_pr_url']}")
-        print(f"   🏠 Repository: {github_config['owner']}/{github_config['repo_name']}")
-
-        agent = GitHubPRAgent()
+        print(f"   🏠 User Fork: {github_config['owner']}/{github_config['repo_name']}")
+        print(f"   🎯 Base Repository: {base_owner}/{base_repo}")
+
+        agent = GitHubPRAgent(
+            user_owner=github_config["owner"],
+            user_repo=github_config["repo_name"],
+            base_owner=base_owner,
+            base_repo=base_repo,
+        )
         result = agent.run_translation_pr_workflow(
             reference_pr_url=github_config["reference_pr_url"],
             target_language=target_language,
             filepath=filepath,
             translated_doc=translated_content,
-            owner=github_config["owner"],
-            repo_name=github_config["repo_name"],
             base_branch=github_config.get("base_branch", "main"),
         )
         # TEST CODE
@@ -202,9 +226,9 @@ def generate_github_pr(
         toctree_result = None
         if en_title:
             from agent.toctree_handler import TocTreeHandler
-            toctree_handler = TocTreeHandler()
+            toctree_handler = TocTreeHandler(project)
             toctree_result = toctree_handler.update_toctree_after_translation(
-                result, filepath, agent, github_config
+                result, filepath, agent, github_config, project
             )
 
         # Process result
@@ -248,13 +272,29 @@ def generate_github_pr(
 {result.get("error_details", "Unknown error")}"""
 
         else:
+            error_details = result.get("error_details", "No additional details")
             return f"""❌ **GitHub PR Creation Failed**
 
 **Error Message:**
-{result["message"]}"""
+{result["message"]}
+
+**Error Details:**
+{error_details}
+
+💡 **Common Solutions:**
+1. **Project Mismatch**: Selected project '{project}' but fork is '{github_config.get('repo_name', 'REPO')}' - ensure they match
+2. Check if your GitHub fork exists: {github_config.get('owner', 'USER')}/{github_config.get('repo_name', 'REPO')}
+3. Verify GitHub token has write access to your fork"""
 
     except Exception as e:
-        error_msg = f"❌ Unexpected error occurred during PR creation: {str(e)}"
+        error_msg = f"""❌ **Unexpected Error During PR Creation**
+
+**Error:** {str(e)}
+
+**Configuration:**
+• Project: {project}
+• File: {filepath}
+• Target: {github_config.get('owner', 'USER')}/{github_config.get('repo_name', 'REPO')} → {base_owner if 'base_owner' in locals() else 'BASE'}/{base_repo if 'base_repo' in locals() else 'REPO'}"""
         print(error_msg)
         return error_msg
 

app.py

@@ -8,17 +8,23 @@ from dotenv import load_dotenv
 
 from agent.handler import (
     approve_handler,
+    confirm_and_go_translate_handler,
+    confirm_translation_and_go_upload_handler,
     get_welcome_message,
     process_file_search_handler,
     restart_handler,
     send_message,
     start_translate_handler,
     sync_language_displays,
+    update_language_selection,
+    update_project_selection,
     update_prompt_preview,
     update_status,
     update_github_config,
+    update_persistent_config,
 )
 from translator.model import Languages
+from translator.project_config import get_available_projects
 
 load_dotenv()
 
@@ -111,12 +117,54 @@ with gr.Blocks(
             gr.Markdown("### 🌐 Hugging Face i18n Agent")
 
             chatbot = gr.Chatbot(
-                value=[[None, get_welcome_message()]], scale=1, height=585,
+                value=[[None, get_welcome_message()]], scale=1, height=525,
                 show_copy_button=True
             )
+            
+            # Chat input directly under main chat
+            gr.Markdown("### 💬 Chat with agent")
+            with gr.Row():
+                msg_input = gr.Textbox(
+                    placeholder="Type your message here... (e.g. 'what', 'how', or 'help')",
+                    container=False,
+                    scale=4,
+                )
+                send_btn = gr.Button("Send", scale=1, elem_classes="action-button")
 
         # Controller interface
         with gr.Column(scale=2):
+            # Configuration Panel
+            with gr.Column(elem_classes=["control-panel"]):
+                gr.Markdown("### ⚙️ Configuration")
+                
+                with gr.Accordion("🔧 API & GitHub Settings", open=True):
+                    config_anthropic_key = gr.Textbox(
+                        label="🔑 Anthropic API Key",
+                        type="password",
+                        placeholder="sk-ant-...",
+                    )
+                    config_github_token = gr.Textbox(
+                        label="🔑 GitHub Token (Required for PR, Optional for file search)",
+                        type="password", 
+                        placeholder="ghp_...",
+                    )
+                    
+                    with gr.Row():
+                        config_github_owner = gr.Textbox(
+                            label="👤 GitHub Owner",
+                            placeholder="your-username",
+                            scale=1,
+                        )
+                        config_github_repo = gr.Textbox(
+                            label="📁 Repository Name", 
+                            placeholder="your-repository",
+                            scale=1,
+                        )
+                    
+                    save_config_btn = gr.Button(
+                        "💾 Save Configuration", elem_classes="action-button"
+                    )
+                    
             # Quick Controller
             with gr.Column(elem_classes=["control-panel"]):
                 gr.Markdown("### 🛠️ Quick Controls")
@@ -125,6 +173,11 @@ with gr.Blocks(
                 with gr.Tabs(elem_classes="simple-tabs") as control_tabs:
                     with gr.TabItem("1. Find Files", id=0):
                         with gr.Group():
+                            project_dropdown = gr.Radio(
+                                choices=get_available_projects(),
+                                label="🎯 Select Project",
+                                value="transformers",
+                            )
                             lang_dropdown = gr.Radio(
                                 choices=[language.value for language in Languages],
                                 label="🌍 Translate To",
@@ -139,6 +192,11 @@ with gr.Blocks(
                                 "🔍 Find Files to Translate",
                                 elem_classes="action-button",
                             )
+                            
+                            confirm_go_btn = gr.Button(
+                                "✅ Confirm Selection & Go to Translate",
+                                elem_classes="action-button",
+                            )
 
                     with gr.TabItem("2. Translate", id=1):
                         with gr.Group():
@@ -159,19 +217,19 @@ with gr.Blocks(
                                 value="ko",
                                 interactive=False,
                             )
-                            anthropic_key = gr.Textbox(
-                                label="🔑 Anthropic API key for translation generation",
-                                type="password",
-                            )
                             additional_instruction = gr.Textbox(
                                 label="📝 Additional instructions (Optional - e.g., custom glossary)",
                                 placeholder="Example: Translate 'model' as '모델' consistently",
                                 lines=2,
                             )
                             
-                            with gr.Accordion("🔍 Preview Prompt", open=False):
+                            force_retranslate = gr.Checkbox(
+                                label="🔄 Force Retranslate (ignore existing translations)",
+                                value=False,
+                            )
+                            
+                            with gr.Accordion("🔍 Preview Translation Prompt", open=False):
                                 prompt_preview = gr.Textbox(
-                                    label="Current Translation Prompt",
                                     lines=8,
                                     interactive=False,
                                     placeholder="Select a file and language to see the prompt preview...",
@@ -181,29 +239,18 @@ with gr.Blocks(
                             start_translate_btn = gr.Button(
                                 "🚀 Start Translation", elem_classes="action-button"
                             )
+                            
+                            confirm_upload_btn = gr.Button(
+                                "✅ Confirm Translation & Upload PR",
+                                elem_classes="action-button",
+                                visible=False,
+                            )
 
                     with gr.TabItem("3. Upload PR", id=2):
                         with gr.Group():
-                            github_token = gr.Textbox(
-                                label="🔑 GitHub Token",
-                                type="password",
-                                placeholder="ghp_xxxxxxxxxxxxxxxxxxxx",
-                            )
-                            github_owner = gr.Textbox(
-                                label="👤 GitHub Owner/Username",
-                                placeholder="your-username",
-                            )
-                            github_repo = gr.Textbox(
-                                label="📁 Repository Name",
-                                placeholder="your-repository",
-                            )
                             reference_pr_url = gr.Textbox(
-                                label="🔗 Reference PR URL (Optional - Agent will find one if not provided)",
-                                placeholder="reference PR URL",
-                            )
-
-                            save_config_btn = gr.Button(
-                                "💾 Save GitHub Config", elem_classes="action-button"
+                                label="🔗 Reference PR URL (Optional)",
+                                placeholder="Auto-filled based on project selection",
                             )
                             approve_btn = gr.Button(
                                 "✅ Generate GitHub PR", elem_classes="action-button"
@@ -212,29 +259,38 @@ with gr.Blocks(
                                 "🔄 Restart Translation", elem_classes="action-button"
                             )
 
-            # Chat Controller
-            with gr.Column(elem_classes=["control-panel"]):
-                gr.Markdown("### 💬 Chat with agent (Only simple chat is available)")
-                msg_input = gr.Textbox(
-                    placeholder="Type your message here... (e.g. 'what', 'how', or 'help')",
-                    container=False,
-                    scale=4,
-                )
-                send_btn = gr.Button("Send", scale=1, elem_classes="action-button")
-
     # Event Handlers
 
     find_btn.click(
         fn=process_file_search_handler,
-        inputs=[lang_dropdown, k_input, chatbot],
+        inputs=[project_dropdown, lang_dropdown, k_input, chatbot],
         outputs=[chatbot, msg_input, status_display, control_tabs, files_to_translate],
     )
+    
+    confirm_go_btn.click(
+        fn=confirm_and_go_translate_handler,
+        inputs=[chatbot],
+        outputs=[chatbot, msg_input, status_display, control_tabs],
+    )
 
-    # Sync language across tabs
+    # Auto-save selections to state and update prompt preview
+    project_dropdown.change(
+        fn=update_project_selection,
+        inputs=[project_dropdown, chatbot],
+        outputs=[chatbot, msg_input, status_display],
+    )
+    
+    # Update prompt preview when project changes
+    project_dropdown.change(
+        fn=update_prompt_preview,
+        inputs=[translate_lang_display, file_to_translate_input, additional_instruction],
+        outputs=[prompt_preview],
+    )
+    
     lang_dropdown.change(
-        fn=sync_language_displays,
-        inputs=[lang_dropdown],
-        outputs=[translate_lang_display],
+        fn=update_language_selection,
+        inputs=[lang_dropdown, chatbot],
+        outputs=[chatbot, msg_input, status_display, translate_lang_display],
     )
 
     #
@@ -247,20 +303,26 @@ with gr.Blocks(
     # Button event handlers
     start_translate_btn.click(
         fn=start_translate_handler,
-        inputs=[chatbot, anthropic_key, file_to_translate_input, additional_instruction],
+        inputs=[chatbot, file_to_translate_input, additional_instruction, force_retranslate],
+        outputs=[chatbot, msg_input, status_display, control_tabs, start_translate_btn, confirm_upload_btn],
+    )
+    
+    confirm_upload_btn.click(
+        fn=confirm_translation_and_go_upload_handler,
+        inputs=[chatbot],
         outputs=[chatbot, msg_input, status_display, control_tabs],
     )
 
-    # GitHub Config Save
+    # Configuration Save
     save_config_btn.click(
-        fn=update_github_config,
-        inputs=[github_token, github_owner, github_repo, reference_pr_url],
-        outputs=[msg_input],
+        fn=update_persistent_config,
+        inputs=[config_anthropic_key, config_github_token, config_github_owner, config_github_repo, reference_pr_url, chatbot],
+        outputs=[chatbot, msg_input, status_display],
     )
 
     approve_btn.click(
         fn=approve_handler,
-        inputs=[chatbot, github_owner, github_repo, reference_pr_url],
+        inputs=[chatbot, config_github_owner, config_github_repo, reference_pr_url],
         outputs=[chatbot, msg_input, status_display],
     )
 

pr_generator/agent.py

@@ -34,9 +34,13 @@ except ImportError as e:
 class GitHubPRAgent:
     """Agent class for GitHub PR creation"""
 
-    def __init__(self):
+    def __init__(self, user_owner: str = None, user_repo: str = None, base_owner: str = None, base_repo: str = None):
         self._github_client = None
         self._llm = None
+        self.user_owner = user_owner
+        self.user_repo = user_repo
+        self.base_owner = base_owner
+        self.base_repo = base_repo
 
     @property
     def github_client(self) -> Optional[Github]:
@@ -433,8 +437,6 @@ Please return only the commit message. No other explanation is needed."""
         target_language: str,
         filepath: str,
         translated_doc: str,
-        owner: str,
-        repo_name: str,
         base_branch: str = "main",
     ) -> Dict[str, Any]:
         """Execute translation document PR creation workflow."""
@@ -458,19 +460,20 @@ Please return only the commit message. No other explanation is needed."""
             )
 
             # 3. Get main branch SHA from upstream and create branch in fork
-            upstream_repo = self.github_client.get_repo(f"huggingface/{repo_name}")
+            upstream_repo = self.github_client.get_repo(f"{self.base_owner}/{self.base_repo}")
             main_branch = upstream_repo.get_branch(base_branch)
             main_sha = main_branch.commit.sha
 
             print(f"🌿 Creating branch: {branch_name} in fork repository")
-            branch_result = self.create_branch(owner, repo_name, branch_name, main_sha)
+            branch_result = self.create_branch(self.user_owner, self.user_repo, branch_name, main_sha)
 
             # Check branch creation result
             if branch_result.startswith("ERROR"):
                 return {
                     "status": "error",
-                    "message": f"Branch creation failed: {branch_result}",
+                    "message": f"Branch creation failed: {branch_result}\n\nTarget: {self.user_owner}/{self.user_repo}\nBranch: {branch_name}\nBase SHA: {main_sha[:8]}",
                     "branch": branch_name,
+                    "error_details": branch_result,
                 }
             elif branch_result.startswith("WARNING"):
                 print(f"⚠️ {branch_result}")
@@ -489,8 +492,8 @@ Please return only the commit message. No other explanation is needed."""
 
             print(f"📄 Saving file: {target_filepath}")
             file_result = self.create_or_update_file(
-                owner,
-                repo_name,
+                self.user_owner,
+                self.user_repo,
                 target_filepath,
                 commit_message,
                 translated_doc,
@@ -500,9 +503,10 @@ Please return only the commit message. No other explanation is needed."""
             if not file_result.startswith("SUCCESS"):
                 return {
                     "status": "error",
-                    "message": "An issue occurred while saving the file.",
+                    "message": f"File save failed: {file_result}\n\n🎯 Target: {self.user_owner}/{self.user_repo} (expected: {target_language} fork of {self.base_owner}/{self.base_repo})\n🌿 Branch: {branch_name}\n📁 File: {target_filepath}",
                     "branch": branch_name,
                     "file_path": target_filepath,
+                    "error_details": file_result,
                 }
 
             print(f"{file_result}")
@@ -518,11 +522,11 @@ Please return only the commit message. No other explanation is needed."""
             )
 
             print(f"🔄 Creating PR: {pr_title}")
-            print(f"   Head: {owner}:{branch_name} → Base: huggingface:{base_branch}")
+            print(f"   Head: {self.user_owner}:{branch_name} → Base: {self.base_owner}:{base_branch}")
 
             # Create PR from fork to upstream repository
             pr_result = self.create_pull_request(
-                "huggingface", "transformers", pr_title, f"{owner}:{branch_name}", base_branch, pr_body, draft=True
+                self.base_owner, self.base_repo, pr_title, f"{self.user_owner}:{branch_name}", base_branch, pr_body, draft=True
             )
 
             if pr_result.startswith("ERROR"):
@@ -554,7 +558,8 @@ Please return only the commit message. No other explanation is needed."""
         except Exception as e:
             return {
                 "status": "error",
-                "message": f"Error occurred during workflow execution: {str(e)}",
+                "message": f"Workflow execution failed: {str(e)}\n\nConfig: {self.user_owner}/{self.user_repo} → {self.base_owner}/{self.base_repo}\nFile: {filepath if 'filepath' in locals() else 'Unknown'}",
+                "error_details": str(e),
             }
 
 

translator/content.py

@@ -6,15 +6,18 @@ from langchain.callbacks import get_openai_callback
 from langchain_anthropic import ChatAnthropic
 
 from translator.prompt_glossary import PROMPT_WITH_GLOSSARY
+from translator.project_config import get_project_config
 
 
-def get_content(filepath: str) -> str:
+def get_content(filepath: str, project: str = "transformers") -> str:
     if filepath == "":
         raise ValueError("No files selected for translation.")
 
-    url = string.Template(
-        "https://raw.githubusercontent.com/huggingface/" "transformers/main/$filepath"
-    ).safe_substitute(filepath=filepath)
+    config = get_project_config(project)
+    # Extract repo path from repo_url (e.g., "huggingface/transformers")
+    repo_path = config.repo_url.replace("https://github.com/", "")
+    
+    url = f"https://raw.githubusercontent.com/{repo_path}/main/{filepath}"
     response = requests.get(url)
     if response.status_code == 200:
         content = response.text
@@ -170,4 +173,4 @@ def llm_translate(to_translate: str) -> tuple[str, str]:
         )
         ai_message = model.invoke(to_translate)
         print("cb:", cb)
-    return cb, ai_message.content
+    return str(cb), ai_message.content

translator/project_config.py

@@ -0,0 +1,48 @@
+"""Project configuration for different HuggingFace repositories."""
+
+from dataclasses import dataclass
+from typing import Dict
+
+
+@dataclass
+class ProjectConfig:
+    """Configuration for a specific HuggingFace project."""
+    name: str
+    repo_url: str
+    api_url: str
+    docs_path: str
+    github_issues: Dict[str, str]  # language -> issue_id
+    reference_pr_url: str
+
+
+# Project configurations
+PROJECTS = {
+    "transformers": ProjectConfig(
+        name="Transformers",
+        repo_url="https://github.com/huggingface/transformers",
+        api_url="https://api.github.com/repos/huggingface/transformers/git/trees/main?recursive=1",
+        docs_path="docs/source",
+        github_issues={"ko": "20179"},
+        reference_pr_url="https://github.com/huggingface/transformers/pull/24968"
+    ),
+    "smolagents": ProjectConfig(
+        name="SmolAgents",
+        repo_url="https://github.com/huggingface/smolagents",
+        api_url="https://api.github.com/repos/huggingface/smolagents/git/trees/main?recursive=1",
+        docs_path="docs/source", 
+        github_issues={"ko": "20179"},  # To be filled when issue is created
+        reference_pr_url="https://github.com/huggingface/smolagents/pull/1581"  # To be filled with actual PR URL
+    )
+}
+
+
+def get_project_config(project_key: str) -> ProjectConfig:
+    """Get project configuration by key."""
+    if project_key not in PROJECTS:
+        raise ValueError(f"Unknown project: {project_key}. Available: {list(PROJECTS.keys())}")
+    return PROJECTS[project_key]
+
+
+def get_available_projects() -> list[str]:
+    """Get list of available project keys."""
+    return list(PROJECTS.keys())

translator/retriever.py

@@ -5,15 +5,26 @@ from pathlib import Path
 import requests
 
 from .model import Languages, Summary, TranslationDoc
+from .project_config import get_project_config
 
-URL = "https://api.github.com/repos/huggingface/transformers/git/trees/main?recursive=1"
 
-
-def get_github_repo_files():
+def get_github_repo_files(project: str = "transformers"):
     """
     Get github repo files
     """
-    response = requests.get(URL)
+    config = get_project_config(project)
+    
+    # Add GitHub token if available to avoid rate limiting (optional)
+    headers = {}
+    github_token = os.environ.get("GITHUB_TOKEN")
+    if github_token:
+        headers["Authorization"] = f"token {github_token}"
+    
+    response = requests.get(config.api_url, headers=headers)
+    
+    # Handle rate limit with helpful message
+    if response.status_code == 403 and "rate limit" in response.text.lower():
+        raise Exception(f"GitHub API rate limit exceeded. To avoid this, set GITHUB_TOKEN in your environment or provide a GitHub token in the UI. Details: {response.text}")
 
     data = response.json()
     all_items = data.get("tree", [])
@@ -26,30 +37,42 @@ def get_github_repo_files():
     return file_paths
 
 
-def get_github_issue_open_pr(lang: str = "ko"):
+def get_github_issue_open_pr(project: str = "transformers", lang: str = "ko", all_files: list = None):
     """
-    Get open PR in the github issue, filtered by title starting with '🌐 [i18n-KO]'.
+    Get open PR in the github issue, filtered by title containing '[i18n-KO]'.
     """
-    if lang == "ko":
-        issue_id = "20179"
-    else:
-        raise ValueError(
-            "No Github issue has been registered to the server. (Only 'ko' is supported - please contact us to support this.)"
-        )
+    config = get_project_config(project)
+    issue_id = config.github_issues.get(lang)
+    
+    # For projects without GitHub issue tracking, still search for PRs
+    if not issue_id:
+        raise ValueError(f"⚠️ No GitHub issue registered for {project}.")
 
+    # Require all_files parameter 
+    if all_files is None:
+        raise ValueError("Repository file list must be provided")
+    
     headers = {
         "Accept": "application/vnd.github+json",
     }
     
+    # Add GitHub token if available to avoid rate limiting (optional)
+    github_token = os.environ.get("GITHUB_TOKEN")
+    if github_token:
+        headers["Authorization"] = f"token {github_token}"
+    
     all_open_prs = []
     page = 1
     per_page = 100  # Maximum allowed by GitHub API
     
     while True:
-        url = f"https://api.github.com/repos/huggingface/transformers/pulls?state=open&page={page}&per_page={per_page}"
+        repo_path = config.repo_url.replace("https://github.com/", "")
+        url = f"https://api.github.com/repos/{repo_path}/pulls?state=open&page={page}&per_page={per_page}"
         response = requests.get(url, headers=headers)
         
-        if response.status_code != 200:
+        if response.status_code == 403 and "rate limit" in response.text.lower():
+            raise Exception(f"GitHub API rate limit exceeded. To avoid this, set GITHUB_TOKEN in your environment or provide a GitHub token in the UI. Details: {response.text}")
+        elif response.status_code != 200:
             raise Exception(f"GitHub API error: {response.status_code} {response.text}")
         
         page_prs = response.json()
@@ -63,19 +86,61 @@ def get_github_issue_open_pr(lang: str = "ko"):
         if len(page_prs) < per_page:
             break
 
-    filtered_prs = [pr for pr in all_open_prs if pr["title"].startswith("🌐 [i18n-KO]")]
+    filtered_prs = [pr for pr in all_open_prs if "[i18n-KO]" in pr["title"]]
 
-    pattern = re.compile(r"`([^`]+\.md)`")
+    # Pattern to match filenames after "Translated" keyword
+    pattern = re.compile(r"Translated\s+(?:`([^`]+)`|(\S+))\s+to")
 
-    filenames = [
-        "docs/source/en/" + match.group(1)
-        for pr in filtered_prs
-        if (match := pattern.search(pr["title"]))
-    ]
-    pr_info_list = [
-        f"https://github.com/huggingface/transformers/pull/{pr['url'].rstrip('/').split('/')[-1]}"
-        for pr in filtered_prs
-    ]
+    def find_original_file_path(filename_from_title, all_files):
+        """Find the exact file path from repo files by matching filename"""
+        if not filename_from_title:
+            return None
+            
+        # Remove .md extension for matching
+        base_name = filename_from_title.replace('.md', '')
+        
+        # Look for exact matches in repo files
+        for file_path in all_files:
+            if file_path.startswith("docs/source/en/") and file_path.endswith(".md"):
+                file_base = file_path.split("/")[-1].replace('.md', '')
+                if file_base == base_name:
+                    return file_path
+                    
+        # If no exact match, fallback to simple path
+        return f"docs/source/en/{filename_from_title}"
+    
+    filenames = []
+    pr_info_list = []
+    
+    for pr in filtered_prs:
+        match = pattern.search(pr["title"])
+        if match:
+            # Use group 1 (with backticks) or group 2 (without backticks)
+            filename = match.group(1) or match.group(2)
+            # Add .md extension if not present
+            if not filename.endswith('.md'):
+                filename += '.md'
+                
+            # Find the correct file path by matching filename
+            correct_path = None
+            if filename:
+                # Remove .md extension for matching
+                base_name = filename.replace('.md', '')
+                
+                # Look for exact matches in repo files
+                for file_path in all_files:
+                    if file_path.startswith("docs/source/en/") and file_path.endswith(".md"):
+                        file_base = file_path.split("/")[-1].replace('.md', '')
+                        if file_base == base_name:
+                            correct_path = file_path
+                            break
+                            
+                # If no exact match, fallback to simple path
+                if not correct_path:
+                    correct_path = f"docs/source/en/{filename}"
+            if correct_path:
+                filenames.append(correct_path)
+                pr_info_list.append(f"{config.repo_url}/pull/{pr['url'].rstrip('/').split('/')[-1]}")
     return filenames, pr_info_list
 
 
@@ -99,11 +164,12 @@ def retrieve(summary: Summary, table_size: int = 10) -> tuple[str, list[str]]:
     return report, first_missing_docs
 
 
-def report(target_lang: str, top_k: int = 1) -> tuple[str, list[str]]:
+def report(project: str, target_lang: str, top_k: int = 1, docs_file: list = None) -> tuple[str, list[str]]:
     """
     Generate a report for the translated docs
     """
-    docs_file = get_github_repo_files()
+    if docs_file is None:
+        raise ValueError("Repository file list must be provided")
 
     base_docs_path = Path("docs/source")
     en_docs_path = Path("docs/source/en")