Skip to content

Azure code interpreter

AzureCodeInterpreterToolSpec #

Bases: BaseToolSpec

Azure Code Interpreter tool spec.

Leverages Azure Dynamic Sessions to execute Python code.

Source code in llama-index-integrations/tools/llama-index-tools-azure-code-interpreter/llama_index/tools/azure_code_interpreter/base.py
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
class AzureCodeInterpreterToolSpec(BaseToolSpec):
    """Azure Code Interpreter tool spec.

    Leverages Azure Dynamic Sessions to execute Python code.
    """

    spec_functions = ["code_interpreter", "list_files"]

    def __init__(
        self,
        pool_management_endpoint: Optional[str] = None,
        session_id: Optional[str] = None,
        local_save_path: Optional[str] = None,
        sanitize_input: bool = True,
    ) -> None:
        """Initialize with parameters."""
        self.pool_management_endpoint: str = pool_management_endpoint or os.getenv(
            "AZURE_POOL_MANAGEMENT_ENDPOINT"
        )
        self.access_token: Optional[AccessToken] = None

        def _access_token_provider_factory() -> Callable[[], Optional[str]]:
            def access_token_provider() -> Optional[str]:
                """Create a function that returns an access token."""
                if self.access_token is None or datetime.fromtimestamp(
                    self.access_token.expires_on, timezone.utc
                ) < (datetime.now(timezone.utc) + timedelta(minutes=5)):
                    credential = DefaultAzureCredential()
                    self.access_token = credential.get_token(
                        "https://dynamicsessions.io/.default"
                    )
                return self.access_token.token

            return access_token_provider

        self.access_token_provider: Callable[
            [], Optional[str]
        ] = _access_token_provider_factory()
        """A function that returns the access token to use for the session pool."""

        self.session_id: str = session_id or str(uuid4())
        """The session ID to use for the session pool. Defaults to a random UUID."""

        self.sanitize_input: bool = sanitize_input
        """Whether to sanitize input before executing it."""

        if local_save_path:
            if not os.path.exists(local_save_path):
                raise Exception(f"Local save path {local_save_path} does not exist.")

        self.local_save_path: Optional[str] = local_save_path
        """The local path to save files generated by Python interpreter."""

        try:
            _package_version = importlib.metadata.version(
                "llamaindex-azure-code-interpreter"
            )
        except importlib.metadata.PackageNotFoundError:
            _package_version = "0.0.0"

        self.user_agent = (
            f"llamaindex-azure-code-interpreter/{_package_version} (Language=Python)"
        )

    def _build_url(self, path: str) -> str:
        pool_management_endpoint = self.pool_management_endpoint
        if not pool_management_endpoint:
            raise ValueError("pool_management_endpoint is not set")

        if not pool_management_endpoint.endswith("/"):
            pool_management_endpoint += "/"

        encoded_session_id = urllib.parse.quote(self.session_id)
        query = f"identifier={encoded_session_id}&api-version=2024-02-02-preview"
        query_separator = "&" if "?" in pool_management_endpoint else "?"

        return pool_management_endpoint + path + query_separator + query

    def code_interpreter(self, python_code: str) -> dict:
        """
        This tool is used to execute python commands when you need to perform calculations or computations in a Session.
        Input should be a valid python command. The tool returns the result, stdout, and stderr.

        Args:
            python_code (str): Python code to be executed generated by llm.
        """
        if self.sanitize_input:
            python_code = _sanitize_input(python_code)

        access_token = self.access_token_provider()
        api_url = self._build_url("code/execute")
        headers = {
            "Authorization": f"Bearer {access_token}",
            "Content-Type": "application/json",
            "User-Agent": self.user_agent,
        }
        body = {
            "properties": {
                "codeInputType": "inline",
                "executionType": "synchronous",
                "code": python_code,
            }
        }

        response = requests.post(api_url, headers=headers, json=body)
        response.raise_for_status()
        response_json = response.json()
        if "properties" in response_json:
            if (
                "result" in response_json["properties"]
                and response_json["properties"]["result"]
            ):
                if isinstance(response_json["properties"]["result"], dict):
                    if "base64_data" in response_json["properties"]["result"]:
                        base64_encoded_data = response_json["properties"]["result"][
                            "base64_data"
                        ]
                        if self.local_save_path:
                            file_path = f"{self.local_save_path}/{self.session_id}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.{response_json['properties']['result']['format']}"
                            decoded_data = base64.b64decode(base64_encoded_data)
                            with open(file_path, "wb") as f:
                                f.write(decoded_data)
                            # Check if file is written to the file path successfully. if so, update the response_json
                            response_json["properties"]["result"][
                                "saved_to_local_path"
                            ] = response_json["properties"]["result"].pop("base64_data")
                            if os.path.exists(file_path):
                                response_json["properties"]["result"][
                                    "saved_to_local_path"
                                ] = True
                            else:
                                response_json["properties"]["result"][
                                    "saved_to_local_path"
                                ] = False
                        else:
                            response_json["properties"]["result"]["base64_data"] = ""
        return response_json

    def upload_file(
        self,
        data: Optional[Any] = None,
        local_file_path: Optional[str] = None,
    ) -> List[RemoteFileMetadata]:
        """Upload a file to the session under the path /mnt/data.

        Args:
            data: The data to upload.
            local_file_path: The path to the local file to upload.

        Returns:
            List[RemoteFileMetadata]: The list of metadatas for the uploaded files.
        """
        if data and local_file_path:
            raise ValueError("data and local_file_path cannot be provided together")

        if local_file_path:
            remote_file_path = f"/mnt/data/{os.path.basename(local_file_path)}"
            data = open(local_file_path, "rb")

        access_token = self.access_token_provider()
        if not remote_file_path.startswith("/mnt/data"):
            remote_file_path = f"/mnt/data/{remote_file_path}"
        api_url = self._build_url("files/upload")
        headers = {
            "Authorization": f"Bearer {access_token}",
        }

        files = [("file", (remote_file_path, data, "application/octet-stream"))]

        response = requests.request("POST", api_url, headers=headers, files=files)
        response.raise_for_status()

        response_json = response.json()
        remote_files_metadatas = []
        for entry in response_json["value"]:
            if "properties" in entry:
                remote_files_metadatas.append(
                    RemoteFileMetadata.from_dict(entry["properties"])
                )
        return remote_files_metadatas

    def download_file_to_local(
        self, remote_file_path: str, local_file_path: Optional[str] = None
    ) -> Optional[BufferedReader]:
        """Download a file from the session back to your local environment.

        Args:
            remote_file_path: The path to download the file from, relative to `/mnt/data`.
            local_file_path: The path to save the downloaded file to. If not provided, the file is returned as a BufferedReader.

        Returns:
            BufferedReader: The data of the downloaded file.
        """
        access_token = self.access_token_provider()
        # In case if the file path LLM provides is absolute, remove the /mnt/data/ prefix
        remote_file_path = remote_file_path.replace("/mnt/data/", "")
        api_url = self._build_url(f"files/content/{remote_file_path}")
        headers = {
            "Authorization": f"Bearer {access_token}",
        }

        response = requests.get(api_url, headers=headers)
        response.raise_for_status()

        if local_file_path:
            with open(local_file_path, "wb") as f:
                f.write(response.content)
            return None

        return BytesIO(response.content)

    def list_files(self) -> List[RemoteFileMetadata]:
        """List the files in the session.

        Returns:
            List[RemoteFileMetadata]: The metadata for the files in the session
        """
        access_token = self.access_token_provider()
        api_url = self._build_url("files")
        headers = {
            "Authorization": f"Bearer {access_token}",
        }

        response = requests.get(api_url, headers=headers)
        response.raise_for_status()

        response_json = response.json()
        return [
            RemoteFileMetadata.from_dict(entry["properties"])
            for entry in response_json["value"]
        ]

access_token_provider instance-attribute #

access_token_provider: Callable[[], Optional[str]] = _access_token_provider_factory()

A function that returns the access token to use for the session pool.

session_id instance-attribute #

session_id: str = session_id or str(uuid4())

The session ID to use for the session pool. Defaults to a random UUID.

sanitize_input instance-attribute #

sanitize_input: bool = sanitize_input

Whether to sanitize input before executing it.

local_save_path instance-attribute #

local_save_path: Optional[str] = local_save_path

The local path to save files generated by Python interpreter.

code_interpreter #

code_interpreter(python_code: str) -> dict

This tool is used to execute python commands when you need to perform calculations or computations in a Session. Input should be a valid python command. The tool returns the result, stdout, and stderr.

Parameters:

Name Type Description Default
python_code str

Python code to be executed generated by llm.

required
Source code in llama-index-integrations/tools/llama-index-tools-azure-code-interpreter/llama_index/tools/azure_code_interpreter/base.py
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
def code_interpreter(self, python_code: str) -> dict:
    """
    This tool is used to execute python commands when you need to perform calculations or computations in a Session.
    Input should be a valid python command. The tool returns the result, stdout, and stderr.

    Args:
        python_code (str): Python code to be executed generated by llm.
    """
    if self.sanitize_input:
        python_code = _sanitize_input(python_code)

    access_token = self.access_token_provider()
    api_url = self._build_url("code/execute")
    headers = {
        "Authorization": f"Bearer {access_token}",
        "Content-Type": "application/json",
        "User-Agent": self.user_agent,
    }
    body = {
        "properties": {
            "codeInputType": "inline",
            "executionType": "synchronous",
            "code": python_code,
        }
    }

    response = requests.post(api_url, headers=headers, json=body)
    response.raise_for_status()
    response_json = response.json()
    if "properties" in response_json:
        if (
            "result" in response_json["properties"]
            and response_json["properties"]["result"]
        ):
            if isinstance(response_json["properties"]["result"], dict):
                if "base64_data" in response_json["properties"]["result"]:
                    base64_encoded_data = response_json["properties"]["result"][
                        "base64_data"
                    ]
                    if self.local_save_path:
                        file_path = f"{self.local_save_path}/{self.session_id}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.{response_json['properties']['result']['format']}"
                        decoded_data = base64.b64decode(base64_encoded_data)
                        with open(file_path, "wb") as f:
                            f.write(decoded_data)
                        # Check if file is written to the file path successfully. if so, update the response_json
                        response_json["properties"]["result"][
                            "saved_to_local_path"
                        ] = response_json["properties"]["result"].pop("base64_data")
                        if os.path.exists(file_path):
                            response_json["properties"]["result"][
                                "saved_to_local_path"
                            ] = True
                        else:
                            response_json["properties"]["result"][
                                "saved_to_local_path"
                            ] = False
                    else:
                        response_json["properties"]["result"]["base64_data"] = ""
    return response_json

upload_file #

upload_file(data: Optional[Any] = None, local_file_path: Optional[str] = None) -> List[RemoteFileMetadata]

Upload a file to the session under the path /mnt/data.

Parameters:

Name Type Description Default
data Optional[Any]

The data to upload.

None
local_file_path Optional[str]

The path to the local file to upload.

None

Returns:

Type Description
List[RemoteFileMetadata]

List[RemoteFileMetadata]: The list of metadatas for the uploaded files.

Source code in llama-index-integrations/tools/llama-index-tools-azure-code-interpreter/llama_index/tools/azure_code_interpreter/base.py
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
def upload_file(
    self,
    data: Optional[Any] = None,
    local_file_path: Optional[str] = None,
) -> List[RemoteFileMetadata]:
    """Upload a file to the session under the path /mnt/data.

    Args:
        data: The data to upload.
        local_file_path: The path to the local file to upload.

    Returns:
        List[RemoteFileMetadata]: The list of metadatas for the uploaded files.
    """
    if data and local_file_path:
        raise ValueError("data and local_file_path cannot be provided together")

    if local_file_path:
        remote_file_path = f"/mnt/data/{os.path.basename(local_file_path)}"
        data = open(local_file_path, "rb")

    access_token = self.access_token_provider()
    if not remote_file_path.startswith("/mnt/data"):
        remote_file_path = f"/mnt/data/{remote_file_path}"
    api_url = self._build_url("files/upload")
    headers = {
        "Authorization": f"Bearer {access_token}",
    }

    files = [("file", (remote_file_path, data, "application/octet-stream"))]

    response = requests.request("POST", api_url, headers=headers, files=files)
    response.raise_for_status()

    response_json = response.json()
    remote_files_metadatas = []
    for entry in response_json["value"]:
        if "properties" in entry:
            remote_files_metadatas.append(
                RemoteFileMetadata.from_dict(entry["properties"])
            )
    return remote_files_metadatas

download_file_to_local #

download_file_to_local(remote_file_path: str, local_file_path: Optional[str] = None) -> Optional[BufferedReader]

Download a file from the session back to your local environment.

Parameters:

Name Type Description Default
remote_file_path str

The path to download the file from, relative to /mnt/data.

required
local_file_path Optional[str]

The path to save the downloaded file to. If not provided, the file is returned as a BufferedReader.

None

Returns:

Name Type Description
BufferedReader Optional[BufferedReader]

The data of the downloaded file.

Source code in llama-index-integrations/tools/llama-index-tools-azure-code-interpreter/llama_index/tools/azure_code_interpreter/base.py
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
def download_file_to_local(
    self, remote_file_path: str, local_file_path: Optional[str] = None
) -> Optional[BufferedReader]:
    """Download a file from the session back to your local environment.

    Args:
        remote_file_path: The path to download the file from, relative to `/mnt/data`.
        local_file_path: The path to save the downloaded file to. If not provided, the file is returned as a BufferedReader.

    Returns:
        BufferedReader: The data of the downloaded file.
    """
    access_token = self.access_token_provider()
    # In case if the file path LLM provides is absolute, remove the /mnt/data/ prefix
    remote_file_path = remote_file_path.replace("/mnt/data/", "")
    api_url = self._build_url(f"files/content/{remote_file_path}")
    headers = {
        "Authorization": f"Bearer {access_token}",
    }

    response = requests.get(api_url, headers=headers)
    response.raise_for_status()

    if local_file_path:
        with open(local_file_path, "wb") as f:
            f.write(response.content)
        return None

    return BytesIO(response.content)

list_files #

list_files() -> List[RemoteFileMetadata]

List the files in the session.

Returns:

Type Description
List[RemoteFileMetadata]

List[RemoteFileMetadata]: The metadata for the files in the session

Source code in llama-index-integrations/tools/llama-index-tools-azure-code-interpreter/llama_index/tools/azure_code_interpreter/base.py
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
def list_files(self) -> List[RemoteFileMetadata]:
    """List the files in the session.

    Returns:
        List[RemoteFileMetadata]: The metadata for the files in the session
    """
    access_token = self.access_token_provider()
    api_url = self._build_url("files")
    headers = {
        "Authorization": f"Bearer {access_token}",
    }

    response = requests.get(api_url, headers=headers)
    response.raise_for_status()

    response_json = response.json()
    return [
        RemoteFileMetadata.from_dict(entry["properties"])
        for entry in response_json["value"]
    ]