Update utils.py

utils.py

@@ -15,9 +15,7 @@ import numpy as np
 import torch
 import random
 
-# --- CHANGE THIS IMPORT ---
-# OLD: from opendeepresearch.agent import OpenDeepResearcher
-# NEW:
+# --- IMPORTANT: USE THIS IMPORT NOW ---
 from smolagents.examples.open_deep_research.agent import OpenDeepResearcher
 from report_structure import generate_report
 
@@ -84,8 +82,8 @@ def query_llm_for_additional_info(topic: str, existing_text: str) -> str:
     # No longer needed
     pass
 def research_topic(topic: str) -> str:
-    # No longer needed
-    pass
+     # No longer needed
+     pass
 
 def fetch_wikipedia_summary(topic: str) -> str:
     # No longer needed
@@ -192,7 +190,7 @@ def generate_script(
             "temperature": 0.7
         }
         response = requests.post("https://openrouter.ai/api/v1/chat/completions",
-                                headers=headers, data=json.dumps(data))
+                                    headers=headers, data=json.dumps(data))
         response.raise_for_status()
         raw_content = response.json()["choices"][0]["message"]["content"].strip()
     except Exception as e:
@@ -222,13 +220,13 @@ def generate_script(
                 d["speaker"] = "Jane"
                 d["display_speaker"] = raw_speaker
 
-        new_dialogue_items = []
-        for d in dialogue_list:
-            if "display_speaker" not in d:
-                d["display_speaker"] = d["speaker"]
-            new_dialogue_items.append(DialogueItem(**d))
+            new_dialogue_items = []
+            for d in dialogue_list:
+                if "display_speaker" not in d:
+                    d["display_speaker"] = d["speaker"]
+                new_dialogue_items.append(DialogueItem(**d))
 
-        return Dialogue(dialogue=new_dialogue_items)
+            return Dialogue(dialogue=new_dialogue_items)
     except json.JSONDecodeError as e:
         print("[ERROR] JSON decoding (format) failed:", e)
         raise ValueError(f"Failed to parse dialogue: {str(e)}")
@@ -236,261 +234,261 @@ def generate_script(
         print("[ERROR] JSON decoding failed:", e)
         raise ValueError(f"Failed to parse dialogue: {str(e)}")
 
-def transcribe_youtube_video(video_url: str) -> str:
-    print("[LOG] Transcribing YouTube video via RapidAPI:", video_url)
-    video_id_match = re.search(r"(?:v=|\/)([0-9A-Za-z_-]{11})", video_url)
-    if not video_id_match:
-        raise ValueError(f"Invalid YouTube URL: {video_url}, cannot extract video ID.")
-
-    video_id = video_id_match.group(1)
-    print("[LOG] Extracted video ID:", video_id)
-
-    base_url = "https://youtube-transcriptor.p.rapidapi.com/transcript"
-    params = {"video_id": video_id, "lang": "en"}
-    headers = {
-        "x-rapidapi-host": "youtube-transcriptor.p.rapidapi.com",
-        "x-rapidapi-key": os.environ.get("RAPIDAPI_KEY")
-    }
-
-    try:
-        response = requests.get(base_url, headers=headers, params=params, timeout=30)
-        print("[LOG] RapidAPI Response Status Code:", response.status_code)
-        print("[LOG] RapidAPI Response Body:", response.text)
-
-        if response.status_code != 200:
-            raise ValueError(f"RapidAPI transcription error: {response.status_code}, {response.text}")
-
-        data = response.json()
-        if not isinstance(data, list) or not data:
-            raise ValueError(f"Unexpected transcript format or empty transcript: {data}")
+    def transcribe_youtube_video(video_url: str) -> str:
+        print("[LOG] Transcribing YouTube video via RapidAPI:", video_url)
+        video_id_match = re.search(r"(?:v=|\/)([0-9A-Za-z_-]{11})", video_url)
+        if not video_id_match:
+            raise ValueError(f"Invalid YouTube URL: {video_url}, cannot extract video ID.")
 
-        transcript_as_text = data[0].get('transcriptionAsText', '').strip()
-        if not transcript_as_text:
-            raise ValueError("transcriptionAsText field is missing or empty.")
+        video_id = video_id_match.group(1)
+        print("[LOG] Extracted video ID:", video_id)
 
-        print("[LOG] Transcript retrieval successful.")
-        print(f"[DEBUG] Transcript Length: {len(transcript_as_text)} characters.")
-        snippet = transcript_as_text[:200] + "..." if len(transcript_as_text) > 200 else transcript_as_text
-        print(f"[DEBUG] Transcript Snippet: {snippet}")
+        base_url = "https://youtube-transcriptor.p.rapidapi.com/transcript"
+        params = {"video_id": video_id, "lang": "en"}
+        headers = {
+            "x-rapidapi-host": "youtube-transcriptor.p.rapidapi.com",
+            "x-rapidapi-key": os.environ.get("RAPIDAPI_KEY")
+        }
 
-        return transcript_as_text
-    except Exception as e:
-        print("[ERROR] RapidAPI transcription error:", e)
-        raise ValueError(f"Error transcribing YouTube video via RapidAPI: {str(e)}")
+        try:
+            response = requests.get(base_url, headers=headers, params=params, timeout=30)
+            print("[LOG] RapidAPI Response Status Code:", response.status_code)
+            print("[LOG] RapidAPI Response Body:", response.text)
 
-def generate_audio_mp3(text: str, speaker: str) -> str:
-    try:
-        import streamlit as st
-        print(f"[LOG] Generating audio for speaker: {speaker}")
-        language_selection = st.session_state.get("language_selection", "English (American)")
-        if language_selection == "English (American)":
-            print(f"[LOG] Using Deepgram for English (American)")
-            if speaker in ["John", "Jane"]:
-                processed_text = text
-            else:
-                processed_text = _preprocess_text_for_tts(text, speaker)
-            deepgram_api_url = "https://api.deepgram.com/v1/speak"
-            params = {"model": "aura-asteria-en"}
-            if speaker == "John":
-                params["model"] = "aura-zeus-en"
-            headers = {
-                "Accept": "audio/mpeg",
-                "Content-Type": "application/json",
-                "Authorization": f"Token {os.environ.get('DEEPGRAM_API_KEY')}"
-            }
-            body = {"text": processed_text}
-            response = requests.post(deepgram_api_url, params=params, headers=headers, json=body, stream=True)
             if response.status_code != 200:
-                raise ValueError(f"Deepgram TTS error: {response.status_code}, {response.text}")
-            content_type = response.headers.get('Content-Type', '')
-            if 'audio/mpeg' not in content_type:
-                raise ValueError("Unexpected Content-Type from Deepgram.")
-            with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as mp3_file:
-                for chunk in response.iter_content(chunk_size=8192):
-                    if chunk:
-                        mp3_file.write(chunk)
-                mp3_path = mp3_file.name
-            audio_seg = AudioSegment.from_file(mp3_path, format="mp3")
-            audio_seg = effects.normalize(audio_seg)
-            final_mp3_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3").name
-            audio_seg.export(final_mp3_path, format="mp3")
-            if os.path.exists(mp3_path):
-                os.remove(mp3_path)
-            return final_mp3_path
+                raise ValueError(f"RapidAPI transcription error: {response.status_code}, {response.text}")
+
+            data = response.json()
+            if not isinstance(data, list) or not data:
+                raise ValueError(f"Unexpected transcript format or empty transcript: {data}")
+
+            transcript_as_text = data[0].get('transcriptionAsText', '').strip()
+            if not transcript_as_text:
+                raise ValueError("transcriptionAsText field is missing or empty.")
+
+            print("[LOG] Transcript retrieval successful.")
+            print(f"[DEBUG] Transcript Length: {len(transcript_as_text)} characters.")
+            snippet = transcript_as_text[:200] + "..." if len(transcript_as_text) > 200 else transcript_as_text
+            print(f"[DEBUG] Transcript Snippet: {snippet}")
+
+            return transcript_as_text
+        except Exception as e:
+            print("[ERROR] RapidAPI transcription error:", e)
+            raise ValueError(f"Error transcribing YouTube video via RapidAPI: {str(e)}")
+
+    def generate_audio_mp3(text: str, speaker: str) -> str:
+        try:
+            import streamlit as st
+            print(f"[LOG] Generating audio for speaker: {speaker}")
+            language_selection = st.session_state.get("language_selection", "English (American)")
+            if language_selection == "English (American)":
+                print(f"[LOG] Using Deepgram for English (American)")
+                if speaker in ["John", "Jane"]:
+                    processed_text = text
+                else:
+                    processed_text = _preprocess_text_for_tts(text, speaker)
+                deepgram_api_url = "https://api.deepgram.com/v1/speak"
+                params = {"model": "aura-asteria-en"}
+                if speaker == "John":
+                    params["model"] = "aura-zeus-en"
+                headers = {
+                    "Accept": "audio/mpeg",
+                    "Content-Type": "application/json",
+                    "Authorization": f"Token {os.environ.get('DEEPGRAM_API_KEY')}"
+                }
+                body = {"text": processed_text}
+                response = requests.post(deepgram_api_url, params=params, headers=headers, json=body, stream=True)
+                if response.status_code != 200:
+                    raise ValueError(f"Deepgram TTS error: {response.status_code}, {response.text}")
+                content_type = response.headers.get('Content-Type', '')
+                if 'audio/mpeg' not in content_type:
+                    raise ValueError("Unexpected Content-Type from Deepgram.")
+                with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as mp3_file:
+                    for chunk in response.iter_content(chunk_size=8192):
+                        if chunk:
+                            mp3_file.write(chunk)
+                    mp3_path = mp3_file.name
+                audio_seg = AudioSegment.from_file(mp3_path, format="mp3")
+                audio_seg = effects.normalize(audio_seg)
+                final_mp3_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3").name
+                audio_seg.export(final_mp3_path, format="mp3")
+                if os.path.exists(mp3_path):
+                    os.remove(mp3_path)
+                return final_mp3_path
+            else:
+                print(f"[LOG] Using Murf API for language: {language_selection}")
+                if language_selection == "Hinglish":
+                    from indic_transliteration.sanscript import transliterate, DEVANAGARI, IAST
+                    text = transliterate(text, DEVANAGARI, IAST)
+                api_key = os.environ.get("MURF_API_KEY")
+                headers = {
+                    "Content-Type": "application/json",
+                    "Accept": "application/json",
+                    "api-key": api_key
+                }
+                multi_native_locale = "hi-IN" if language_selection in ["Hinglish", "Hindi"] else "en-IN"
+                if language_selection == "English (Indian)":
+                    voice_id = "en-IN-aarav" if speaker == "John" else "en-IN-isha"
+                elif language_selection == "Hindi":
+                    voice_id = "hi-IN-kabir" if speaker == "John" else "hi-IN-shweta"
+                elif language_selection == "Hinglish":
+                    voice_id = "hi-IN-kabir" if speaker == "John" else "hi-IN-shweta"
+                else:
+                    voice_id = "en-IN-aarav" if speaker == "John" else "en-IN-isha"
+                payload = {
+                    "audioDuration": 0,
+                    "channelType": "MONO",
+                    "encodeAsBase64": False,
+                    "format": "WAV",
+                    "modelVersion": "GEN2",
+                    "multiNativeLocale": multi_native_locale,
+                    "pitch": 0,
+                    "pronunciationDictionary": {},
+                    "rate": 0,
+                    "sampleRate": 48000,
+                    "style": "Conversational",
+                    "text": text,
+                    "variation": 1,
+                    "voiceId": voice_id
+                }
+                response = requests.post("https://api.murf.ai/v1/speech/generate", headers=headers, json=payload)
+                if response.status_code != 200:
+                    raise ValueError(f"Murf API error: {response.status_code}, {response.text}")
+                json_resp = response.json()
+                audio_url = json_resp.get("audioFile")
+                if not audio_url:
+                    raise ValueError("No audio file URL returned by Murf API")
+                audio_response = requests.get(audio_url)
+                if audio_response.status_code != 200:
+                    raise ValueError(f"Error fetching audio from {audio_url}")
+                with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as wav_file:
+                    wav_file.write(audio_response.content)
+                    wav_path = wav_file.name
+                audio_seg = AudioSegment.from_file(wav_path, format="wav")
+                audio_seg = effects.normalize(audio_seg)
+                final_mp3_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3").name
+                audio_seg.export(final_mp3_path, format="mp3")
+                os.remove(wav_path)
+                return final_mp3_path
+        except Exception as e:
+            print("[ERROR] Error generating audio:", e)
+            raise ValueError(f"Error generating audio: {str(e)}")
+
+    def transcribe_youtube_video_OLD_YTDLP(video_url: str) -> str:
+        pass
+
+    def _preprocess_text_for_tts(text: str, speaker: str) -> str:
+        text = re.sub(r"\bNo\.\b", "Number", text)
+        text = re.sub(r"\b(?i)SaaS\b", "sass", text)
+        abbreviations_as_words = {"NASA", "NATO", "UNESCO"}
+        def insert_periods_for_abbrev(m):
+            abbr = m.group(0)
+            if abbr in abbreviations_as_words:
+                return abbr
+            return ".".join(list(abbr)) + "."
+        text = re.sub(r"\b([A-Z]{2,})\b", insert_periods_for_abbrev, text)
+        text = re.sub(r"\.\.", ".", text)
+        def remove_periods_for_tts(m):
+            return m.group().replace(".", " ").strip()
+        text = re.sub(r"[A-Z]\.[A-Z](?:\.[A-Z])*\.", remove_periods_for_tts, text)
+        text = re.sub(r"-", " ", text)
+        text = re.sub(r"\b(ha(ha)?|heh|lol)\b", "(* laughs *)", text, flags=re.IGNORECASE)
+        text = re.sub(r"\bsigh\b", "(* sighs *)", text, flags=re.IGNORECASE)
+        text = re.sub(r"\b(groan|moan)\b", "(* groans *)", text, flags=re.IGNORECASE)
+        if speaker != "Jane":
+            def insert_thinking_pause(m):
+                word = m.group(1)
+                if random.random() < 0.3:
+                    filler = random.choice(['hmm,', 'well,', 'let me see,'])
+                    return f"{word}..., {filler}"
+                else:
+                    return f"{word}...,"
+            keywords_pattern = r"\b(important|significant|crucial|point|topic)\b"
+            text = re.sub(keywords_pattern, insert_thinking_pause, text, flags=re.IGNORECASE)
+            conj_pattern = r"\b(and|but|so|because|however)\b"
+            text = re.sub(conj_pattern, lambda m: f"{m.group()}...", text, flags=re.IGNORECASE)
+        text = re.sub(r"\b(uh|um|ah)\b", "", text, flags=re.IGNORECASE)
+        def capitalize_match(m):
+            return m.group().upper()
+        text = re.sub(r'(^\s*\w)|([.!?]\s*\w)', capitalize_match, text)
+        return text.strip()
+
+    def _spell_digits(d: str) -> str:
+        digit_map = {
+            '0': 'zero', '1': 'one', '2': 'two', '3': 'three',
+            '4': 'four', '5': 'five', '6': 'six', '7': 'seven',
+            '8': 'eight', '9': 'nine'
+        }
+        return " ".join(digit_map[ch] for ch in d if ch in digit_map)
+
+    def mix_with_bg_music(spoken: AudioSegment, custom_music_path=None) -> AudioSegment:
+        if custom_music_path:
+            music_path = custom_music_path
         else:
-            print(f"[LOG] Using Murf API for language: {language_selection}")
-            if language_selection == "Hinglish":
-                from indic_transliteration.sanscript import transliterate, DEVANAGARI, IAST
-                text = transliterate(text, DEVANAGARI, IAST)
-            api_key = os.environ.get("MURF_API_KEY")
+            music_path = "bg_music.mp3"
+
+        try:
+            bg_music = AudioSegment.from_file(music_path, format="mp3")
+        except Exception as e:
+            print("[ERROR] Failed to load background music:", e)
+            return spoken
+
+        bg_music = bg_music - 18.0
+        total_length_ms = len(spoken) + 2000
+        looped_music = AudioSegment.empty()
+        while len(looped_music) < total_length_ms:
+            looped_music += bg_music
+        looped_music = looped_music[:total_length_ms]
+        final_mix = looped_music.overlay(spoken, position=2000)
+        return final_mix
+
+    def call_groq_api_for_qa(system_prompt: str) -> str:
+        #Kept for use, Changed model
+        try:
             headers = {
+                "Authorization": f"Bearer {os.environ.get('GROQ_API_KEY')}",  # Use GROQ API KEY
                 "Content-Type": "application/json",
-                "Accept": "application/json",
-                "api-key": api_key
+                "Accept": "application/json"
             }
-            multi_native_locale = "hi-IN" if language_selection in ["Hinglish", "Hindi"] else "en-IN"
-            if language_selection == "English (Indian)":
-                voice_id = "en-IN-aarav" if speaker == "John" else "en-IN-isha"
-            elif language_selection == "Hindi":
-                voice_id = "hi-IN-kabir" if speaker == "John" else "hi-IN-shweta"
-            elif language_selection == "Hinglish":
-                voice_id = "hi-IN-kabir" if speaker == "John" else "hi-IN-shweta"
-            else:
-                voice_id = "en-IN-aarav" if speaker == "John" else "en-IN-isha"
-            payload = {
-                "audioDuration": 0,
-                "channelType": "MONO",
-                "encodeAsBase64": False,
-                "format": "WAV",
-                "modelVersion": "GEN2",
-                "multiNativeLocale": multi_native_locale,
-                "pitch": 0,
-                "pronunciationDictionary": {},
-                "rate": 0,
-                "sampleRate": 48000,
-                "style": "Conversational",
-                "text": text,
-                "variation": 1,
-                "voiceId": voice_id
+            data = {
+                "model": "deepseek-r1-distill-llama-70b", #Using Deepseek
+                "messages": [{"role": "user", "content": system_prompt}],
+                "max_tokens": 512,
+                "temperature": 0.7
             }
-            response = requests.post("https://api.murf.ai/v1/speech/generate", headers=headers, json=payload)
-            if response.status_code != 200:
-                raise ValueError(f"Murf API error: {response.status_code}, {response.text}")
-            json_resp = response.json()
-            audio_url = json_resp.get("audioFile")
-            if not audio_url:
-                raise ValueError("No audio file URL returned by Murf API")
-            audio_response = requests.get(audio_url)
-            if audio_response.status_code != 200:
-                raise ValueError(f"Error fetching audio from {audio_url}")
-            with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as wav_file:
-                wav_file.write(audio_response.content)
-                wav_path = wav_file.name
-            audio_seg = AudioSegment.from_file(wav_path, format="wav")
-            audio_seg = effects.normalize(audio_seg)
-            final_mp3_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3").name
-            audio_seg.export(final_mp3_path, format="mp3")
-            os.remove(wav_path)
-            return final_mp3_path
-    except Exception as e:
-        print("[ERROR] Error generating audio:", e)
-        raise ValueError(f"Error generating audio: {str(e)}")
-
-def transcribe_youtube_video_OLD_YTDLP(video_url: str) -> str:
-    pass
-
-def _preprocess_text_for_tts(text: str, speaker: str) -> str:
-    text = re.sub(r"\bNo\.\b", "Number", text)
-    text = re.sub(r"\b(?i)SaaS\b", "sass", text)
-    abbreviations_as_words = {"NASA", "NATO", "UNESCO"}
-    def insert_periods_for_abbrev(m):
-        abbr = m.group(0)
-        if abbr in abbreviations_as_words:
-            return abbr
-        return ".".join(list(abbr)) + "."
-    text = re.sub(r"\b([A-Z]{2,})\b", insert_periods_for_abbrev, text)
-    text = re.sub(r"\.\.", ".", text)
-    def remove_periods_for_tts(m):
-        return m.group().replace(".", " ").strip()
-    text = re.sub(r"[A-Z]\.[A-Z](?:\.[A-Z])*\.", remove_periods_for_tts, text)
-    text = re.sub(r"-", " ", text)
-    text = re.sub(r"\b(ha(ha)?|heh|lol)\b", "(* laughs *)", text, flags=re.IGNORECASE)
-    text = re.sub(r"\bsigh\b", "(* sighs *)", text, flags=re.IGNORECASE)
-    text = re.sub(r"\b(groan|moan)\b", "(* groans *)", text, flags=re.IGNORECASE)
-    if speaker != "Jane":
-        def insert_thinking_pause(m):
-            word = m.group(1)
-            if random.random() < 0.3:
-                filler = random.choice(['hmm,', 'well,', 'let me see,'])
-                return f"{word}..., {filler}"
-            else:
-                return f"{word}...,"
-        keywords_pattern = r"\b(important|significant|crucial|point|topic)\b"
-        text = re.sub(keywords_pattern, insert_thinking_pause, text, flags=re.IGNORECASE)
-        conj_pattern = r"\b(and|but|so|because|however)\b"
-        text = re.sub(conj_pattern, lambda m: f"{m.group()}...", text, flags=re.IGNORECASE)
-    text = re.sub(r"\b(uh|um|ah)\b", "", text, flags=re.IGNORECASE)
-    def capitalize_match(m):
-        return m.group().upper()
-    text = re.sub(r'(^\s*\w)|([.!?]\s*\w)', capitalize_match, text)
-    return text.strip()
-
-def _spell_digits(d: str) -> str:
-    digit_map = {
-        '0': 'zero', '1': 'one', '2': 'two', '3': 'three',
-        '4': 'four', '5': 'five', '6': 'six', '7': 'seven',
-        '8': 'eight', '9': 'nine'
-    }
-    return " ".join(digit_map[ch] for ch in d if ch in digit_map)
-
-def mix_with_bg_music(spoken: AudioSegment, custom_music_path=None) -> AudioSegment:
-    if custom_music_path:
-        music_path = custom_music_path
-    else:
-        music_path = "bg_music.mp3"
-
-    try:
-        bg_music = AudioSegment.from_file(music_path, format="mp3")
-    except Exception as e:
-        print("[ERROR] Failed to load background music:", e)
-        return spoken
-
-    bg_music = bg_music - 18.0
-    total_length_ms = len(spoken) + 2000
-    looped_music = AudioSegment.empty()
-    while len(looped_music) < total_length_ms:
-        looped_music += bg_music
-    looped_music = looped_music[:total_length_ms]
-    final_mix = looped_music.overlay(spoken, position=2000)
-    return final_mix
-
-def call_groq_api_for_qa(system_prompt: str) -> str:
-    #Kept for use, Changed model
-    try:
-        headers = {
-            "Authorization": f"Bearer {os.environ.get('GROQ_API_KEY')}",  # Use GROQ API KEY
-            "Content-Type": "application/json",
-            "Accept": "application/json"
-        }
-        data = {
-            "model": "deepseek-r1-distill-llama-70b", #Using Deepseek
-            "messages": [{"role": "user", "content": system_prompt}],
-            "max_tokens": 512,
-            "temperature": 0.7
-        }
-        response = requests.post("https://api.groq.com/openai/v1/chat/completions", #Using groq endpoint
-                                headers=headers, data=json.dumps(data))
-        response.raise_for_status()
-        return response.json()["choices"][0]["message"]["content"].strip()
-    except Exception as e:
-        print("[ERROR] Groq API error:", e)
-        fallback = {"speaker": "John", "text": "I'm sorry, I'm having trouble answering right now."}
-        return json.dumps(fallback)
-
-# --- Agent and Tavily Integration ---
-
-def run_research_agent(topic: str, report_type: str = "research_report", max_results: int = 20) -> str:
-    """
-    Runs the Open Deep Research agent to generate a research report.
-
-    Args:
-        topic: The research topic.
-        report_type: The type of report to generate (currently only supports "research_report").
-        max_results: The maximum number of search results to use.
-
-    Returns:
-        A string containing the generated research report.  Or, in case of error,
-        an error message.
-    """
-    print(f"[LOG] Starting research agent for topic: {topic}")
-    try:
-        agent = OpenDeepResearcher(topic, report_type=report_type, max_results=max_results, tavily_api_key=os.environ.get("TAVILY_API_KEY"))
-        report_content = agent.run()
-        print("[LOG] Research agent completed successfully.")
-
-        # Now, use the report_structure module to generate the structured report.
-        structured_report = generate_report(report_content)
-        return structured_report
-    except Exception as e:
-        print(f"[ERROR] Error in research agent: {e}")
-        return f"Sorry, I encountered an error during research: {e}"
+            response = requests.post("https://api.groq.com/openai/v1/chat/completions", #Using groq endpoint
+                                    headers=headers, data=json.dumps(data))
+            response.raise_for_status()
+            return response.json()["choices"][0]["message"]["content"].strip()
+        except Exception as e:
+            print("[ERROR] Groq API error:", e)
+            fallback = {"speaker": "John", "text": "I'm sorry, I'm having trouble answering right now."}
+            return json.dumps(fallback)
+
+    # --- Agent and Tavily Integration ---
+
+    def run_research_agent(topic: str, report_type: str = "research_report", max_results: int = 20) -> str:
+        """
+        Runs the Open Deep Research agent to generate a research report.
+
+        Args:
+            topic: The research topic.
+            report_type: The type of report to generate (currently only supports "research_report").
+            max_results: The maximum number of search results to use.
+
+        Returns:
+            A string containing the generated research report.  Or, in case of error,
+            an error message.
+        """
+        print(f"[LOG] Starting research agent for topic: {topic}")
+        try:
+            agent = OpenDeepResearcher(topic, report_type=report_type, max_results=max_results, tavily_api_key=os.environ.get("TAVILY_API_KEY"))
+            report_content = agent.run()
+            print("[LOG] Research agent completed successfully.")
+
+            # Now, use the report_structure module to generate the structured report.
+            structured_report = generate_report(report_content)
+            return structured_report
+        except Exception as e:
+            print(f"[ERROR] Error in research agent: {e}")
+            return f"Sorry, I encountered an error during research: {e}"