first commit

backend/app/gpt/base.py

@@ -0,0 +1,13 @@
+from abc import ABC,abstractmethod
+
+from app.models.gpt_model import GPTSource
+
+
+class GPT(ABC):
+    def summarize(self, source:GPTSource )->str:
+        '''
+
+        :param source: 
+        :return:
+        '''
+        pass

backend/app/gpt/deepseek_gpt.py

@@ -0,0 +1,59 @@
+from typing import List
+from app.gpt.base import GPT
+from openai import OpenAI
+from app.gpt.prompt import BASE_PROMPT, AI_SUM, SCREENSHOT
+from app.gpt.utils import fix_markdown
+from app.models.gpt_model import GPTSource
+from app.models.transcriber_model import TranscriptSegment
+from datetime import timedelta
+
+
+class DeepSeekGPT(GPT):
+    def __init__(self):
+        from os import getenv
+        self.api_key = getenv("DEEP_SEEK_API_KEY")
+        self.base_url = getenv("DEEP_SEEK_API_BASE_URL")
+        self.model=getenv('DEEP_SEEK_MODEL')
+        print(self.model)
+        self.client = OpenAI(api_key=self.api_key, base_url=self.base_url)
+        self.screenshot = False
+
+    def _format_time(self, seconds: float) -> str:
+        return str(timedelta(seconds=int(seconds)))[2:]  # e.g., 03:15
+
+    def _build_segment_text(self, segments: List[TranscriptSegment]) -> str:
+        return "\n".join(
+            f"{self._format_time(seg.start)} - {seg.text.strip()}"
+            for seg in segments
+        )
+
+    def ensure_segments_type(self, segments) -> List[TranscriptSegment]:
+        return [
+            TranscriptSegment(**seg) if isinstance(seg, dict) else seg
+            for seg in segments
+        ]
+
+    def create_messages(self, segments: List[TranscriptSegment], title: str,tags:str):
+        content = BASE_PROMPT.format(
+            video_title=title,
+            segment_text=self._build_segment_text(segments),
+            tags=tags
+        )
+        if self.screenshot:
+            print(":需要截图")
+            content += SCREENSHOT
+        print(content)
+        return [{"role": "user", "content": content + AI_SUM}]
+
+    def summarize(self, source: GPTSource) -> str:
+        self.screenshot = source.screenshot
+        source.segment = self.ensure_segments_type(source.segment)
+        messages = self.create_messages(source.segment, source.title,source.tags)
+        response = self.client.chat.completions.create(
+            model=self.model,
+            messages=messages,
+            temperature=0.7
+        )
+        return response.choices[0].message.content.strip()
+
+

backend/app/gpt/openai_gpt.py

@@ -0,0 +1,65 @@
+from typing import List
+from app.gpt.base import GPT
+from openai import OpenAI
+from app.gpt.prompt import BASE_PROMPT, AI_SUM, SCREENSHOT, LINK
+from app.gpt.utils import fix_markdown
+from app.models.gpt_model import GPTSource
+from app.models.transcriber_model import TranscriptSegment
+from datetime import timedelta
+
+
+class OpenaiGPT(GPT):
+    def __init__(self):
+        from os import getenv
+        self.api_key = getenv("OPENAI_API_KEY")
+        self.base_url = getenv("OPENAI_API_BASE_URL")
+        self.model=getenv('OPENAI_MODEL')
+        print(self.model)
+        self.client = OpenAI(api_key=self.api_key, base_url=self.base_url)
+        self.screenshot = False
+        self.link=False
+
+    def _format_time(self, seconds: float) -> str:
+        return str(timedelta(seconds=int(seconds)))[2:]  # e.g., 03:15
+
+    def _build_segment_text(self, segments: List[TranscriptSegment]) -> str:
+        return "\n".join(
+            f"{self._format_time(seg.start)} - {seg.text.strip()}"
+            for seg in segments
+        )
+
+    def ensure_segments_type(self, segments) -> List[TranscriptSegment]:
+        return [
+            TranscriptSegment(**seg) if isinstance(seg, dict) else seg
+            for seg in segments
+        ]
+
+    def create_messages(self, segments: List[TranscriptSegment], title: str,tags:str):
+        content = BASE_PROMPT.format(
+            video_title=title,
+            segment_text=self._build_segment_text(segments),
+            tags=tags
+        )
+        if self.screenshot:
+            print(":需要截图")
+            content += SCREENSHOT
+        if self.link:
+            print(":需要链接")
+            content += LINK
+
+        print(content)
+        return [{"role": "user", "content": content + AI_SUM}]
+
+    def summarize(self, source: GPTSource) -> str:
+        self.screenshot = source.screenshot
+        self.link = source.link
+        source.segment = self.ensure_segments_type(source.segment)
+        messages = self.create_messages(source.segment, source.title,source.tags)
+        response = self.client.chat.completions.create(
+            model=self.model,
+            messages=messages,
+            temperature=0.7
+        )
+        return response.choices[0].message.content.strip()
+
+

backend/app/gpt/prompt.py

@@ -0,0 +1,56 @@
+BASE_PROMPT = '''
+You are a professional note-taking assistant who excels at summarizing video transcripts into clear, structured, and information-rich notes.
+
+🎯 Language Requirement:
+- The notes must be written in **Chinese**.
+- Proper nouns, technical terms, brand names, and personal names should remain in **English** where appropriate.
+
+📌 Video Title:
+{video_title}
+
+📎 Video Tags:
+{tags}
+
+📝 Your Task:
+Based on the segmented transcript below, generate structured notes in standard **Markdown format**, and follow these principles:
+
+1. **Complete information**: Record as much relevant detail as possible to ensure comprehensive coverage.
+2. **Clear structure**: Organize content with logical sectioning. Use appropriate heading levels (`##`, `###`) to summarize key points in each section.
+3. **Concise wording**: Use accurate, clear, and professional Chinese expressions.
+4. **Remove irrelevant content**: Omit advertisements, filler words, casual greetings, and off-topic remarks.
+5. **Keep critical details**: Preserve important facts, examples, conclusions, and recommendations.
+6. **Readable layout**: Use bullet points where needed, and keep paragraphs reasonably short to enhance readability.
+7. **Table of Contents**: Generate a table of contents at the top based on the `##` level headings.
+
+
+⚠️ Output Instructions:
+- Only return the final **Markdown content**.
+- Do **not** wrap the output in code blocks like ```` ```markdown ```` or ```` ``` ````.
+
+
+🎬 Transcript Segments (Format: Start Time - Text):
+
+---
+{segment_text}
+---
+'''
+
+LINK='''
+9. **Add time markers**: THIS IS IMPORTANT For every main heading (`##`), append the starting time of that segment using the format ,start with *Content ,eg: `*Content-[mm:ss]`.
+
+
+'''
+AI_SUM='''
+
+🧠 Final Touch:
+At the end of the notes, add a professional **AI Summary** in Chinese – a brief conclusion summarizing the whole video.
+
+
+
+'''
+
+SCREENSHOT='''
+8. **Screenshot placeholders**: If a section involves **visual demonstrations, code walkthroughs, UI interactions**, or any content where visuals aid understanding, insert a screenshot cue at the end of that section:
+   - Format: `*Screenshot-[mm:ss]`
+   - Only use it when truly helpful.
+'''

backend/app/gpt/qwen_gpt.py

@@ -0,0 +1,59 @@
+from typing import List
+from app.gpt.base import GPT
+from openai import OpenAI
+from app.gpt.prompt import BASE_PROMPT, AI_SUM, SCREENSHOT
+from app.gpt.utils import fix_markdown
+from app.models.gpt_model import GPTSource
+from app.models.transcriber_model import TranscriptSegment
+from datetime import timedelta
+
+
+class QwenGPT(GPT):
+    def __init__(self):
+        from os import getenv
+        self.api_key = getenv("QWEN_API_KEY")
+        self.base_url = getenv("QWEN_API_BASE_URL")
+        self.model=getenv('QWEN_MODEL')
+        print(self.model)
+        self.client = OpenAI(api_key=self.api_key, base_url=self.base_url)
+        self.screenshot = False
+
+    def _format_time(self, seconds: float) -> str:
+        return str(timedelta(seconds=int(seconds)))[2:]  # e.g., 03:15
+
+    def _build_segment_text(self, segments: List[TranscriptSegment]) -> str:
+        return "\n".join(
+            f"{self._format_time(seg.start)} - {seg.text.strip()}"
+            for seg in segments
+        )
+
+    def ensure_segments_type(self, segments) -> List[TranscriptSegment]:
+        return [
+            TranscriptSegment(**seg) if isinstance(seg, dict) else seg
+            for seg in segments
+        ]
+
+    def create_messages(self, segments: List[TranscriptSegment], title: str,tags:str):
+        content = BASE_PROMPT.format(
+            video_title=title,
+            segment_text=self._build_segment_text(segments),
+            tags=tags
+        )
+        if self.screenshot:
+            print(":需要截图")
+            content += SCREENSHOT
+        print(content)
+        return [{"role": "user", "content": content + AI_SUM}]
+
+    def summarize(self, source: GPTSource) -> str:
+        self.screenshot = source.screenshot
+        source.segment = self.ensure_segments_type(source.segment)
+        messages = self.create_messages(source.segment, source.title,source.tags)
+        response = self.client.chat.completions.create(
+            model=self.model,
+            messages=messages,
+            temperature=0.7
+        )
+        return response.choices[0].message.content.strip()
+
+

backend/app/gpt/utils.py

@@ -0,0 +1,4 @@
+import codecs
+
+def fix_markdown(markdown: str) -> str:
+    return codecs.decode(markdown, 'unicode_escape')