Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -27,10 +27,6 @@ def count_tokens(text: str, model: str) -> int:
|
|
| 27 |
return len(enc.encode(text))
|
| 28 |
|
| 29 |
def trim_conversation(convo: list[dict], model: str, max_context: int) -> list[dict]:
|
| 30 |
-
"""
|
| 31 |
-
Slide-window: drop the oldest user/assistant turns until
|
| 32 |
-
total tokens + BUFFER_TOKENS β€ max_context.
|
| 33 |
-
"""
|
| 34 |
tokens = [count_tokens(m["content"], model) for m in convo]
|
| 35 |
total = sum(tokens)
|
| 36 |
while total + BUFFER_TOKENS > max_context and len(convo) > 2:
|
|
@@ -41,11 +37,6 @@ def trim_conversation(convo: list[dict], model: str, max_context: int) -> list[d
|
|
| 41 |
return convo
|
| 42 |
|
| 43 |
def safe_chat(convo: list[dict], max_context: int, max_reply: int, model_list: list[str]):
|
| 44 |
-
"""
|
| 45 |
-
1) Try each model in model_list in order.
|
| 46 |
-
2) On model-not-found: skip to next.
|
| 47 |
-
3) On context-length error: trim history & retry same model once.
|
| 48 |
-
"""
|
| 49 |
last_exc = None
|
| 50 |
for model in model_list:
|
| 51 |
try:
|
|
@@ -57,11 +48,9 @@ def safe_chat(convo: list[dict], max_context: int, max_reply: int, model_list: l
|
|
| 57 |
)
|
| 58 |
except Exception as e:
|
| 59 |
msg = str(e).lower()
|
| 60 |
-
# model-not-found β fallback
|
| 61 |
if "does not exist" in msg or "model_not_found" in msg or "404" in msg:
|
| 62 |
last_exc = e
|
| 63 |
continue
|
| 64 |
-
# context-length β trim & retry
|
| 65 |
if "context length" in msg or "maximum context length" in msg:
|
| 66 |
trimmed = trim_conversation(convo.copy(), model, max_context)
|
| 67 |
try:
|
|
@@ -74,9 +63,7 @@ def safe_chat(convo: list[dict], max_context: int, max_reply: int, model_list: l
|
|
| 74 |
except Exception as e2:
|
| 75 |
last_exc = e2
|
| 76 |
continue
|
| 77 |
-
# other errors β bubble up
|
| 78 |
raise
|
| 79 |
-
# none succeeded
|
| 80 |
raise last_exc or RuntimeError("All models failed in safe_chat()")
|
| 81 |
|
| 82 |
def chat_handler(
|
|
@@ -87,20 +74,17 @@ def chat_handler(
|
|
| 87 |
max_context: int,
|
| 88 |
max_reply: int
|
| 89 |
) -> tuple[list[tuple[str, str]], str]:
|
| 90 |
-
"""Gradio handler: builds convo, calls safe_chat, updates history."""
|
| 91 |
if not user_message.strip():
|
| 92 |
return history, ""
|
| 93 |
if not client.api_key:
|
| 94 |
return history, "β OPENAI_API_KEY not set."
|
| 95 |
|
| 96 |
-
# Build conversation payload
|
| 97 |
convo = [{"role":"system","content":system_prompt}]
|
| 98 |
for u, b in history or []:
|
| 99 |
convo.append({"role":"user", "content":u})
|
| 100 |
convo.append({"role":"assistant", "content":b})
|
| 101 |
convo.append({"role":"user","content":user_message})
|
| 102 |
|
| 103 |
-
# Prepare model list: user choice first, then the rest
|
| 104 |
fallback = [m for m in ALL_MODELS if m != selected_model]
|
| 105 |
models_to_try = [selected_model] + fallback
|
| 106 |
|
|
@@ -115,18 +99,13 @@ def chat_handler(
|
|
| 115 |
return history, ""
|
| 116 |
|
| 117 |
def clear_chat_handler() -> list:
|
| 118 |
-
"""Clear all previous chat messages."""
|
| 119 |
return []
|
| 120 |
|
| 121 |
# βββ Gradio UI βββ
|
| 122 |
-
with gr.Blocks(
|
| 123 |
-
title="π€ CodeBot: Your Cool Coding Assistant"
|
| 124 |
-
) as demo:
|
| 125 |
-
|
| 126 |
gr.Markdown(
|
| 127 |
"""
|
| 128 |
## CodeBot
|
| 129 |
-
A stylish coding assistant with:
|
| 130 |
- **Model selector**: pick your LLM
|
| 131 |
- **Context slider**: control how much history to keep
|
| 132 |
- **Reply slider**: set max response length
|
|
@@ -153,27 +132,17 @@ with gr.Blocks(
|
|
| 153 |
label="βοΈ Max Reply Tokens"
|
| 154 |
)
|
| 155 |
|
| 156 |
-
# Example questions
|
| 157 |
examples = [
|
| 158 |
"How do I implement quicksort in Python?",
|
| 159 |
"Show me a C# example using LINQ to group items.",
|
| 160 |
"Explain async/await in Python with sample code.",
|
| 161 |
"How to connect to SQL Server using C#?"
|
| 162 |
]
|
|
|
|
| 163 |
with gr.Row():
|
| 164 |
-
example_dropdown = gr.Dropdown(
|
| 165 |
-
choices=examples,
|
| 166 |
-
label="π‘ Examples"
|
| 167 |
-
)
|
| 168 |
example_btn = gr.Button("π₯ Load Example")
|
| 169 |
|
| 170 |
-
# Prefill user_input when an example is selected
|
| 171 |
-
example_btn.click(
|
| 172 |
-
fn=lambda q: q or "",
|
| 173 |
-
inputs=[example_dropdown],
|
| 174 |
-
outputs=[gr.update(component_id="user_input")]
|
| 175 |
-
)
|
| 176 |
-
|
| 177 |
system_txt = gr.Textbox(
|
| 178 |
lines=3,
|
| 179 |
value=(
|
|
@@ -194,6 +163,13 @@ with gr.Blocks(
|
|
| 194 |
elem_id="user_input"
|
| 195 |
)
|
| 196 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 197 |
with gr.Row():
|
| 198 |
send_btn = gr.Button("π Send")
|
| 199 |
clear_btn = gr.Button("ποΈ Clear Chat", variant="secondary")
|
|
|
|
| 27 |
return len(enc.encode(text))
|
| 28 |
|
| 29 |
def trim_conversation(convo: list[dict], model: str, max_context: int) -> list[dict]:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
tokens = [count_tokens(m["content"], model) for m in convo]
|
| 31 |
total = sum(tokens)
|
| 32 |
while total + BUFFER_TOKENS > max_context and len(convo) > 2:
|
|
|
|
| 37 |
return convo
|
| 38 |
|
| 39 |
def safe_chat(convo: list[dict], max_context: int, max_reply: int, model_list: list[str]):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
last_exc = None
|
| 41 |
for model in model_list:
|
| 42 |
try:
|
|
|
|
| 48 |
)
|
| 49 |
except Exception as e:
|
| 50 |
msg = str(e).lower()
|
|
|
|
| 51 |
if "does not exist" in msg or "model_not_found" in msg or "404" in msg:
|
| 52 |
last_exc = e
|
| 53 |
continue
|
|
|
|
| 54 |
if "context length" in msg or "maximum context length" in msg:
|
| 55 |
trimmed = trim_conversation(convo.copy(), model, max_context)
|
| 56 |
try:
|
|
|
|
| 63 |
except Exception as e2:
|
| 64 |
last_exc = e2
|
| 65 |
continue
|
|
|
|
| 66 |
raise
|
|
|
|
| 67 |
raise last_exc or RuntimeError("All models failed in safe_chat()")
|
| 68 |
|
| 69 |
def chat_handler(
|
|
|
|
| 74 |
max_context: int,
|
| 75 |
max_reply: int
|
| 76 |
) -> tuple[list[tuple[str, str]], str]:
|
|
|
|
| 77 |
if not user_message.strip():
|
| 78 |
return history, ""
|
| 79 |
if not client.api_key:
|
| 80 |
return history, "β OPENAI_API_KEY not set."
|
| 81 |
|
|
|
|
| 82 |
convo = [{"role":"system","content":system_prompt}]
|
| 83 |
for u, b in history or []:
|
| 84 |
convo.append({"role":"user", "content":u})
|
| 85 |
convo.append({"role":"assistant", "content":b})
|
| 86 |
convo.append({"role":"user","content":user_message})
|
| 87 |
|
|
|
|
| 88 |
fallback = [m for m in ALL_MODELS if m != selected_model]
|
| 89 |
models_to_try = [selected_model] + fallback
|
| 90 |
|
|
|
|
| 99 |
return history, ""
|
| 100 |
|
| 101 |
def clear_chat_handler() -> list:
|
|
|
|
| 102 |
return []
|
| 103 |
|
| 104 |
# βββ Gradio UI βββ
|
| 105 |
+
with gr.Blocks(title="π€ CodeBot: Your Cool Coding Assistant") as demo:
|
|
|
|
|
|
|
|
|
|
| 106 |
gr.Markdown(
|
| 107 |
"""
|
| 108 |
## CodeBot
|
|
|
|
| 109 |
- **Model selector**: pick your LLM
|
| 110 |
- **Context slider**: control how much history to keep
|
| 111 |
- **Reply slider**: set max response length
|
|
|
|
| 132 |
label="βοΈ Max Reply Tokens"
|
| 133 |
)
|
| 134 |
|
|
|
|
| 135 |
examples = [
|
| 136 |
"How do I implement quicksort in Python?",
|
| 137 |
"Show me a C# example using LINQ to group items.",
|
| 138 |
"Explain async/await in Python with sample code.",
|
| 139 |
"How to connect to SQL Server using C#?"
|
| 140 |
]
|
| 141 |
+
|
| 142 |
with gr.Row():
|
| 143 |
+
example_dropdown = gr.Dropdown(choices=examples, label="π‘ Examples")
|
|
|
|
|
|
|
|
|
|
| 144 |
example_btn = gr.Button("π₯ Load Example")
|
| 145 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 146 |
system_txt = gr.Textbox(
|
| 147 |
lines=3,
|
| 148 |
value=(
|
|
|
|
| 163 |
elem_id="user_input"
|
| 164 |
)
|
| 165 |
|
| 166 |
+
# Wire up example loader to the actual user_input component
|
| 167 |
+
example_btn.click(
|
| 168 |
+
fn=lambda q: q or "",
|
| 169 |
+
inputs=[example_dropdown],
|
| 170 |
+
outputs=[user_input]
|
| 171 |
+
)
|
| 172 |
+
|
| 173 |
with gr.Row():
|
| 174 |
send_btn = gr.Button("π Send")
|
| 175 |
clear_btn = gr.Button("ποΈ Clear Chat", variant="secondary")
|