refactor: update content limits for quiz and summary generation modules

quiz_generator/constants.py

@@ -8,6 +8,13 @@ MAX_TF_COUNT = 20
 MAX_SAMPLE_CHUNKS = 10
 RETRIEVER_K = 5
 
+# Web search configuration
+WIKI_TOP_K_RESULTS = 2
+WIKI_DOC_CONTENT_CHARS_MAX = 15000
+
+ARXIV_TOP_K_RESULTS = 4
+ARXIV_DOC_CONTENT_CHARS_MAX = 10000
+
 QUIZ_PROMPT_TEMPLATE = PromptTemplate(
     input_variables=[
         "difficulty", "mcq_count", "tf_count",

quiz_generator/quiz.py

@@ -11,6 +11,10 @@ from .constants import (
     QUIZ_PROMPT_TEMPLATE,
     MAX_SAMPLE_CHUNKS,
     RETRIEVER_K,
+    WIKI_TOP_K_RESULTS,
+    WIKI_DOC_CONTENT_CHARS_MAX,
+    ARXIV_TOP_K_RESULTS,
+    ARXIV_DOC_CONTENT_CHARS_MAX,
 )
 
 logger = logging.getLogger(__name__)
@@ -121,7 +125,12 @@ def _web_quiz(difficulty, mcq_count, tf_count, topic_title):
     try:
         prompt = _quiz_prompt()
         llm = get_llm()
-        tools = web_search_tools()
+        tools = web_search_tools(
+            wiki_k=WIKI_TOP_K_RESULTS,
+            wiki_chars=WIKI_DOC_CONTENT_CHARS_MAX,
+            arxiv_k=ARXIV_TOP_K_RESULTS,
+            arxiv_chars=ARXIV_DOC_CONTENT_CHARS_MAX,
+        )
         agent = create_tool_calling_agent(llm, tools, prompt)
 
         executor = AgentExecutor(

rag/constants.py

@@ -4,6 +4,13 @@ BATCH_MAX_SIZE = 8
 BATCH_WINDOW_S = 0.05
 WARMUP_INTERVAL_S = 300
 
+# Web search configuration
+WIKI_TOP_K_RESULTS = 2
+WIKI_DOC_CONTENT_CHARS_MAX = 3000
+ARXIV_TOP_K_RESULTS = 3
+ARXIV_DOC_CONTENT_CHARS_MAX = 2500
+DUCKDUCKGO_DOC_CONTENT_CHARS_MAX = 3000
+
 RAG_PROMPT_TEMPLATE_BASE = """\
 <role>
 You are a helpful AI study assistant. You provide accurate, well-reasoned educational answers.{subject_line}

rag/rag.py

@@ -25,6 +25,11 @@ from .constants import (
     EMBEDDING_DIM,
     RAG_PROMPT_TEMPLATE_BASE,
     CHAT_TITLE_PROMPT_TEMPLATE,
+    WIKI_TOP_K_RESULTS,
+    WIKI_DOC_CONTENT_CHARS_MAX,
+    ARXIV_TOP_K_RESULTS,
+    ARXIV_DOC_CONTENT_CHARS_MAX,
+    DUCKDUCKGO_DOC_CONTENT_CHARS_MAX,
 )
 from .schemas import SearchInput, EmbeddingJob
 
@@ -159,14 +164,16 @@ def get_groq_llm():
 
 # ── Web Search Tools ───────────────────────────────────
 
-def web_search_tools(has_material: bool = False):
+def web_search_tools(
+    has_material: bool = False,
+    wiki_k: int = WIKI_TOP_K_RESULTS,
+    wiki_chars: int = WIKI_DOC_CONTENT_CHARS_MAX,
+    arxiv_k: int = ARXIV_TOP_K_RESULTS,
+    arxiv_chars: int = ARXIV_DOC_CONTENT_CHARS_MAX,
+    duck_chars: int = DUCKDUCKGO_DOC_CONTENT_CHARS_MAX,
+):
 
     tools = []
-    
-    # Target ~16,500 chars total across all tools
-    wiki_k = 2;   wiki_chars = 3000    # 6,000 total
-    arxiv_k = 3;  arxiv_chars = 2500   # 7,500 total
-    duck_chars = 3000                   # 3,000 total
 
     try:
         wiki_api = WikipediaAPIWrapper(top_k_results=wiki_k, doc_content_chars_max=wiki_chars)

summary_generator/constants.py

@@ -1,7 +1,14 @@
 from langchain.prompts import PromptTemplate
 
 MAX_INPUT_CHARS = 15000
-MAX_COMBINED_TEXT_LEN = 80000
+MAX_COMBINED_TEXT_LEN = 160000
+
+# Web search configuration
+WIKI_TOP_K_RESULTS = 2
+WIKI_DOC_CONTENT_CHARS_MAX = 40000  
+
+ARXIV_TOP_K_RESULTS = 3
+ARXIV_DOC_CONTENT_CHARS_MAX = 25000 
 
 SUMMARIZER_PROMPT_TEMPLATE = PromptTemplate(
     input_variables=["input"],

summary_generator/routes.py

@@ -44,7 +44,8 @@ async def generate_summary(
                 raise HTTPException(400, "No text chunks found in this material")
             combined = "\n".join(c["content"] for c in chunks_list)
             if len(combined) > MAX_COMBINED_TEXT_LEN:
-                combined = combined[:MAX_COMBINED_TEXT_LEN]
+                half_len = MAX_COMBINED_TEXT_LEN // 2
+                combined = combined[:half_len] + combined[-half_len:]
             summary = await loop.run_in_executor(None, summarizer, combined)
 
         elapsed = time.time() - start

summary_generator/summary.py

@@ -2,7 +2,14 @@ import re
 import logging
 from langchain_community.utilities import ArxivAPIWrapper, WikipediaAPIWrapper
 from src.rag.rag import get_llm
-from .constants import SUMMARIZER_PROMPT_TEMPLATE, MAX_INPUT_CHARS
+from .constants import (
+    SUMMARIZER_PROMPT_TEMPLATE,
+    MAX_INPUT_CHARS,
+    WIKI_TOP_K_RESULTS,
+    WIKI_DOC_CONTENT_CHARS_MAX,
+    ARXIV_TOP_K_RESULTS,
+    ARXIV_DOC_CONTENT_CHARS_MAX,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -67,7 +74,10 @@ def web_summarizer(topic: str) -> str:
     all_content = []
 
     try:
-        wiki_api = WikipediaAPIWrapper(top_k_results=2, doc_content_chars_max=6000)
+        wiki_api = WikipediaAPIWrapper(
+            top_k_results=WIKI_TOP_K_RESULTS,
+            doc_content_chars_max=WIKI_DOC_CONTENT_CHARS_MAX,
+        )
         wiki_content = wiki_api.run(topic)
         if wiki_content and wiki_content.strip():
             all_content.append(f"--- Wikipedia ---\n{wiki_content}")
@@ -75,7 +85,10 @@ def web_summarizer(topic: str) -> str:
         logger.warning(f"Wikipedia search for '{topic}' failed: {e}")
 
     try:
-        arxiv_api = ArxivAPIWrapper(top_k_results=1, doc_content_chars_max=7000)
+        arxiv_api = ArxivAPIWrapper(
+            top_k_results=ARXIV_TOP_K_RESULTS,
+            doc_content_chars_max=ARXIV_DOC_CONTENT_CHARS_MAX,
+        )
         arxiv_content = arxiv_api.run(topic)
         if arxiv_content and arxiv_content.strip():
             all_content.append(f"--- Arxiv ---\n{arxiv_content}")