Update agent.py
Browse files
agent.py
CHANGED
|
@@ -134,35 +134,35 @@ def gaia_file_reader(file_id: str) -> str:
|
|
| 134 |
# Final agent class
|
| 135 |
# --------------------------------------------------------------------------- #
|
| 136 |
class GeminiAgent:
|
| 137 |
-
"""
|
| 138 |
-
Exposed to `app.py` – instantiated once and then called per question.
|
| 139 |
-
"""
|
| 140 |
-
|
| 141 |
def __init__(self):
|
| 142 |
-
model = GeminiModel()
|
| 143 |
-
tools = [
|
| 144 |
-
PythonInterpreterTool(), # maths, csv, small image ops
|
| 145 |
-
DuckDuckGoSearchTool(), # quick web look-ups
|
| 146 |
-
gaia_file_reader, # our custom file tool
|
| 147 |
-
]
|
| 148 |
self.system_prompt = (
|
| 149 |
"You are a concise, highly accurate assistant. "
|
| 150 |
"Unless explicitly required, reply with ONE short sentence. "
|
| 151 |
"Use the provided tools if needed. "
|
| 152 |
"All answers are graded by exact string match."
|
| 153 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
self.agent = CodeAgent(
|
| 155 |
model=model,
|
| 156 |
tools=tools,
|
| 157 |
-
|
|
|
|
| 158 |
)
|
| 159 |
-
print("✅ GeminiAgent
|
| 160 |
|
| 161 |
-
# ----------- Main entry point for app.py ----------- #
|
| 162 |
def __call__(self, question: str) -> str:
|
| 163 |
file_ids = FILE_TAG.findall(question)
|
|
|
|
|
|
|
| 164 |
if file_ids:
|
| 165 |
-
# Build multimodal user content
|
| 166 |
parts: list[gtypes.Part] = []
|
| 167 |
text_part = FILE_TAG.sub("", question).strip()
|
| 168 |
if text_part:
|
|
@@ -171,15 +171,23 @@ class GeminiAgent:
|
|
| 171 |
try:
|
| 172 |
img_bytes = _download_file(fid)
|
| 173 |
mime = mimetypes.guess_type(fid)[0] or "image/png"
|
| 174 |
-
parts.append(
|
|
|
|
|
|
|
| 175 |
except Exception as exc:
|
| 176 |
-
parts.append(
|
|
|
|
|
|
|
| 177 |
messages = [
|
| 178 |
{"role": "system", "content": self.system_prompt},
|
| 179 |
{"role": "user", "content": parts},
|
| 180 |
]
|
| 181 |
answer = self.agent.model.call_messages(messages)
|
|
|
|
|
|
|
| 182 |
else:
|
| 183 |
-
|
| 184 |
-
|
|
|
|
|
|
|
| 185 |
return answer.rstrip(" .\n\r\t")
|
|
|
|
| 134 |
# Final agent class
|
| 135 |
# --------------------------------------------------------------------------- #
|
| 136 |
class GeminiAgent:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 137 |
def __init__(self):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138 |
self.system_prompt = (
|
| 139 |
"You are a concise, highly accurate assistant. "
|
| 140 |
"Unless explicitly required, reply with ONE short sentence. "
|
| 141 |
"Use the provided tools if needed. "
|
| 142 |
"All answers are graded by exact string match."
|
| 143 |
)
|
| 144 |
+
|
| 145 |
+
model = GeminiModel()
|
| 146 |
+
tools = [
|
| 147 |
+
PythonInterpreterTool(),
|
| 148 |
+
DuckDuckGoSearchTool(),
|
| 149 |
+
gaia_file_reader,
|
| 150 |
+
]
|
| 151 |
+
|
| 152 |
+
# ✨ system_prompt removed – newest smolagents doesn't take it
|
| 153 |
self.agent = CodeAgent(
|
| 154 |
model=model,
|
| 155 |
tools=tools,
|
| 156 |
+
# any other kwargs (executor_type, additional_authorized_imports…)
|
| 157 |
+
verbosity_level=0,
|
| 158 |
)
|
| 159 |
+
print("✅ GeminiAgent ready.")
|
| 160 |
|
|
|
|
| 161 |
def __call__(self, question: str) -> str:
|
| 162 |
file_ids = FILE_TAG.findall(question)
|
| 163 |
+
|
| 164 |
+
# -------- multimodal branch -------- #
|
| 165 |
if file_ids:
|
|
|
|
| 166 |
parts: list[gtypes.Part] = []
|
| 167 |
text_part = FILE_TAG.sub("", question).strip()
|
| 168 |
if text_part:
|
|
|
|
| 171 |
try:
|
| 172 |
img_bytes = _download_file(fid)
|
| 173 |
mime = mimetypes.guess_type(fid)[0] or "image/png"
|
| 174 |
+
parts.append(
|
| 175 |
+
gtypes.Part.from_bytes(data=img_bytes, mime_type=mime)
|
| 176 |
+
)
|
| 177 |
except Exception as exc:
|
| 178 |
+
parts.append(
|
| 179 |
+
gtypes.Part.from_text(f"[FILE {fid} ERROR: {exc}]")
|
| 180 |
+
)
|
| 181 |
messages = [
|
| 182 |
{"role": "system", "content": self.system_prompt},
|
| 183 |
{"role": "user", "content": parts},
|
| 184 |
]
|
| 185 |
answer = self.agent.model.call_messages(messages)
|
| 186 |
+
|
| 187 |
+
# -------- text-only branch -------- #
|
| 188 |
else:
|
| 189 |
+
# prepend system prompt to the user question
|
| 190 |
+
full_prompt = f"{self.system_prompt}\n\n{question}"
|
| 191 |
+
answer = self.agent(full_prompt)
|
| 192 |
+
|
| 193 |
return answer.rstrip(" .\n\r\t")
|