Skip to content

Inference

  • Just run ./start_local.sh and it will take care of everything.
  • The UI should either pop up or you can navigate to http://localhost:8501/ in your browser.
  • Note that it takes a decent bit of time to load everything.

Stopping

  • Run ./stop_local.sh
  • ./start_local.sh stores the PIDs of all the processes it starts in files in all the directories it starts them in. stop_local.sh reads these files and kills the processes.

CLI access to the API

  • We all are lazy sometimes and don't want to use the interface sometimes. Or just want to test out different parts of the API without any hassle. To that end, you can either test out the individual components like so:
  • Note that the %20 are spaces in the URL.

Ollama

  • This is the server that runs an Ollama server (This is basically an optimized version of a local LLM. It does not do anything of itself but runs as a background service so you can use the LLM).
  • You can start it by running cd ollama && ./get_ollama.sh &

LLM Service

  • This component is the one that runs the query processing using LLMs module. It uses the Ollama server, runs queries and processes them.
  • You can start it by running cd llm_service && uvicorn llm_service:app --host 0.0.0.0 --port 8081 &
  • Curl Example : curl http://0.0.0.0:8081/llmquery/find%20me%20a%20mushroom%20dataset%20with%20less%20than%203000%20classes

Backend

  • This component runs the RAG pipeline. It returns a JSON with dataset ids of the OpenML datasets that match the query.
  • You can start it by running cd backend && uvicorn backend:app --host 0.0.0.0 --port 8000 &
  • Curl Example : curl http://0.0.0.0:8000/dataset/find%20me%20a%20mushroom%20dataset

Frontend

  • This component runs the Streamlit frontend. It is the UI that you see when you navigate to http://localhost:8501.
  • You can start it by running cd frontend && streamlit run ui.py &

Errors

  • If you get an error about file permissions, run chmod +x start_local.sh and chmod +x stop_local.sh to make them executable.

LLMResponseParser

Description: Parse the response from the LLM service and update the columns based on the response.

Source code in frontend/ui_utils.py 
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
class LLMResponseParser:
    """
    Description: Parse the response from the LLM service and update the columns based on the response.
    """

    def __init__(self, llm_response):
        self.llm_response = llm_response
        self.subset_cols = ["did", "name"]
        self.size_sort = None
        self.classification_type = None
        self.uploader_name = None

    def process_size_attribute(self, attr_size: str):
        size, sort = attr_size.split(",") if "," in attr_size else (attr_size, None)
        if size == "yes":
            self.subset_cols.append("NumberOfInstances")
        if sort:
            self.size_sort = sort

    def missing_values_attribute(self, attr_missing: str):
        if attr_missing == "yes":
            self.subset_cols.append("NumberOfMissingValues")

    def classification_type_attribute(self, attr_classification: str):
        if attr_classification != "none":
            self.subset_cols.append("NumberOfClasses")
            self.classification_type = attr_classification

    def uploader_attribute(self, attr_uploader: str):
        if attr_uploader != "none":
            self.subset_cols.append("uploader")
            self.uploader_name = attr_uploader.split("=")[1].strip()

    def get_attributes_from_response(self):
        attribute_processors = {
            "size_of_dataset": self.process_size_attribute,
            "missing_values": self.missing_values_attribute,
            "classification_type": self.classification_type_attribute,
            "uploader": self.uploader_attribute,
        }

        for attribute, value in self.llm_response.items():
            if attribute in attribute_processors:
                attribute_processors[attribute](value)

    def update_subset_cols(self, metadata: pd.DataFrame):
        """
        Description: Filter the metadata based on the updated subset columns and extra conditions
        """
        if self.classification_type is not None:
            if "multi" in self.classification_type:
                metadata = metadata[metadata["NumberOfClasses"] > 2]
            elif "binary" in self.classification_type:
                metadata = metadata[metadata["NumberOfClasses"] == 2]
        if self.uploader_name is not None:
            try:
                uploader = int(self.uploader_name)
                metadata = metadata[metadata["uploader"] == uploader]
            except:
                pass

        return metadata[self.subset_cols]

update_subset_cols(metadata)

Description: Filter the metadata based on the updated subset columns and extra conditions

Source code in frontend/ui_utils.py 
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
def update_subset_cols(self, metadata: pd.DataFrame):
    """
    Description: Filter the metadata based on the updated subset columns and extra conditions
    """
    if self.classification_type is not None:
        if "multi" in self.classification_type:
            metadata = metadata[metadata["NumberOfClasses"] > 2]
        elif "binary" in self.classification_type:
            metadata = metadata[metadata["NumberOfClasses"] == 2]
    if self.uploader_name is not None:
        try:
            uploader = int(self.uploader_name)
            metadata = metadata[metadata["uploader"] == uploader]
        except:
            pass

    return metadata[self.subset_cols]

ResponseParser

Description : This classe is used to decide the order of operations and run the response parsing. It loads the paths, fetches the Query parsing LLM response, the rag response, loads the metadatas and then based on the config, decides the order in which to apply each of them.

Source code in frontend/ui_utils.py 
 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
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
class ResponseParser:
    """
    Description : This classe is used to decide the order of operations and run the response parsing.
    It loads the paths, fetches the Query parsing LLM response, the rag response, loads the metadatas and then based on the config, decides the order in which to apply each of them.
    """

    def __init__(self, query_type: str, apply_llm_before_rag: bool = False):
        self.query_type = query_type
        self.paths = self.load_paths()
        self.rag_response = None
        self.llm_response = None
        self.apply_llm_before_rag = apply_llm_before_rag
        self.database_filtered = None
        self.structured_query_response = None

    def load_paths(self):
        """
        Description: Load paths from paths.json
        """
        with open("paths.json", "r") as file:
            return json.load(file)

    def fetch_llm_response(self, query: str):
        """
        Description: Fetch the response from the query parsing LLM service as a json
        """
        llm_response_path = self.paths["llm_response"]
        try:
            self.llm_response = requests.get(
                f"{llm_response_path['docker']}{query}"
            ).json()
        except:
            self.llm_response = requests.get(
                f"{llm_response_path['local']}{query}"
            ).json()
        return self.llm_response

    def fetch_structured_query(self, query_type: str, query: str):
        """
        Description: Fetch the response for a structured query from the LLM service as a JSON
        """
        structured_response_path = self.paths["structured_query"]
        try:
            self.structured_query_response = requests.get(
                f"{structured_response_path['docker']}{query}",
                json={"query": query},
            ).json()
        except (requests.exceptions.RequestException, json.JSONDecodeError) as e:
            # Print the error for debugging purposes
            print(f"Error occurred: {e}")
            # Set structured_query_response to None on error
            self.structured_query_response = None
        try:
            self.structured_query_response = requests.get(
                f"{structured_response_path['local']}{query}",
                json={"query": query},
            ).json()
        except Exception as e:
            # Print the error for debugging purposes
            print(f"Error occurred while fetching from local endpoint: {e}")
            # Set structured_query_response to None if the local request also fails
            self.structured_query_response = None

        return self.structured_query_response

    def database_filter(self, filter_condition, collec):
        """
        Apply database filter on the rag_response
        """
        ids = list(map(str, self.rag_response["initial_response"]))
        self.database_filtered = collec.get(ids=ids, where=filter_condition)["ids"]
        self.database_filtered = list(map(int, self.database_filtered))
        # print(self.database_filtered)
        return self.database_filtered

    def fetch_rag_response(self, query_type, query):
        """
        Description: Fetch the response from RAG pipeline

        """
        rag_response_path = self.paths["rag_response"]
        try:
            self.rag_response = requests.get(
                f"{rag_response_path['docker']}{query_type.lower()}/{query}",
                json={"query": query, "type": query_type.lower()},
            ).json()
        except:
            self.rag_response = requests.get(
                f"{rag_response_path['local']}{query_type.lower()}/{query}",
                json={"query": query, "type": query_type.lower()},
            ).json()
        ordered_set = self._order_results()
        self.rag_response["initial_response"] = ordered_set

        return self.rag_response

    def _order_results(self):
        doc_set = set()
        ordered_set = []
        for docid in self.rag_response["initial_response"]:
            if docid not in doc_set:
                ordered_set.append(docid)
            doc_set.add(docid)
        return ordered_set

    def parse_and_update_response(self, metadata: pd.DataFrame):
        """
         Description: Parse the response from the RAG and LLM services and update the metadata based on the response.
         Decide which order to apply them
         -  self.apply_llm_before_rag == False
             - Metadata is filtered based on the rag response first and then by the Query parsing LLM
        -  self.apply_llm_before_rag == False
             - Metadata is filtered based by the Query parsing LLM first and the rag response second
        - in case structured_query == true, take results are applying data filters.
        """
        if self.apply_llm_before_rag is None or self.llm_response is None:
            print("No LLM filter.")
            # print(self.rag_response, flush=True)
            filtered_metadata = self._no_filter(metadata)

            # print(filtered_metadata)
            # if no llm response is required, return the initial response
            return filtered_metadata

        elif (
            self.rag_response is not None and self.llm_response is not None
        ) and not config["structured_query"]:
            if not self.apply_llm_before_rag:
                filtered_metadata, llm_parser = self._rag_before_llm(metadata)

                if self.query_type.lower() == "dataset":
                    llm_parser.get_attributes_from_response()
                    return llm_parser.update_subset_cols(filtered_metadata)

            elif self.apply_llm_before_rag:
                filtered_metadata = self._filter_before_rag(metadata)
                return filtered_metadata

        elif (
            self.rag_response is not None and self.structured_query_response is not None
        ):
            col_name = [
                "status",
                "NumberOfClasses",
                "NumberOfFeatures",
                "NumberOfInstances",
            ]
            # print(self.structured_query_response)  # Only for debugging. Comment later.
            if self.structured_query_response[0] is not None and isinstance(
                self.structured_query_response[1], dict
            ):
                # Safely attempt to access the "filter" key in the first element

                self._structured_query_on_success(metadata)

            else:
                filtered_metadata = self._structured_query_on_fail(metadata)
                # print("Showing only rag response")
            return filtered_metadata[["did", "name", *col_name]]

    def _structured_query_on_fail(self, metadata):
        filtered_metadata = metadata[
            metadata["did"].isin(self.rag_response["initial_response"])
        ]
        filtered_metadata["did"] = pd.Categorical(
            filtered_metadata["did"],
            categories=self.rag_response["initial_response"],
            ordered=True,
        )
        filtered_metadata = filtered_metadata.sort_values("did").reset_index(drop=True)

        return filtered_metadata

    def _structured_query_on_success(self, metadata):
        if (
            self.structured_query_response[0].get("filter", None)
            and self.database_filtered
        ):
            filtered_metadata = metadata[metadata["did"].isin(self.database_filtered)]
            # print("Showing database filtered data")
        else:
            filtered_metadata = metadata[
                metadata["did"].isin(self.rag_response["initial_response"])
            ]
            # print(
            #     "Showing only rag response as filter is empty or none of the rag data satisfies filter conditions."
            # )
        filtered_metadata["did"] = pd.Categorical(
            filtered_metadata["did"],
            categories=self.rag_response["initial_response"],
            ordered=True,
        )
        filtered_metadata = filtered_metadata.sort_values("did").reset_index(drop=True)

    def _filter_before_rag(self, metadata):
        print("LLM filter before RAG")
        llm_parser = LLMResponseParser(self.llm_response)
        llm_parser.get_attributes_from_response()
        filtered_metadata = llm_parser.update_subset_cols(metadata)
        filtered_metadata = filtered_metadata[
            metadata["did"].isin(self.rag_response["initial_response"])
        ]
        filtered_metadata["did"] = pd.Categorical(
            filtered_metadata["did"],
            categories=self.rag_response["initial_response"],
            ordered=True,
        )
        filtered_metadata = filtered_metadata.sort_values("did").reset_index(drop=True)

        return filtered_metadata

    def _rag_before_llm(self, metadata):
        print("RAG before LLM filter.")
        filtered_metadata = metadata[
            metadata["did"].isin(self.rag_response["initial_response"])
        ]
        filtered_metadata["did"] = pd.Categorical(
            filtered_metadata["did"],
            categories=self.rag_response["initial_response"],
            ordered=True,
        )
        filtered_metadata = filtered_metadata.sort_values("did").reset_index(drop=True)
        llm_parser = LLMResponseParser(self.llm_response)
        return filtered_metadata, llm_parser

    def _no_filter(self, metadata):
        filtered_metadata = metadata[
            metadata["did"].isin(self.rag_response["initial_response"])
        ]
        filtered_metadata["did"] = pd.Categorical(
            filtered_metadata["did"],
            categories=self.rag_response["initial_response"],
            ordered=True,
        )
        filtered_metadata = filtered_metadata.sort_values("did").reset_index(drop=True)

        return filtered_metadata

database_filter(filter_condition, collec)

Apply database filter on the rag_response

Source code in frontend/ui_utils.py 
143
144
145
146
147
148
149
150
151
def database_filter(self, filter_condition, collec):
    """
    Apply database filter on the rag_response
    """
    ids = list(map(str, self.rag_response["initial_response"]))
    self.database_filtered = collec.get(ids=ids, where=filter_condition)["ids"]
    self.database_filtered = list(map(int, self.database_filtered))
    # print(self.database_filtered)
    return self.database_filtered

fetch_llm_response(query)

Description: Fetch the response from the query parsing LLM service as a json

Source code in frontend/ui_utils.py 
100
101
102
103
104
105
106
107
108
109
110
111
112
113
def fetch_llm_response(self, query: str):
    """
    Description: Fetch the response from the query parsing LLM service as a json
    """
    llm_response_path = self.paths["llm_response"]
    try:
        self.llm_response = requests.get(
            f"{llm_response_path['docker']}{query}"
        ).json()
    except:
        self.llm_response = requests.get(
            f"{llm_response_path['local']}{query}"
        ).json()
    return self.llm_response

fetch_rag_response(query_type, query)

Description: Fetch the response from RAG pipeline

Source code in frontend/ui_utils.py 
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
def fetch_rag_response(self, query_type, query):
    """
    Description: Fetch the response from RAG pipeline

    """
    rag_response_path = self.paths["rag_response"]
    try:
        self.rag_response = requests.get(
            f"{rag_response_path['docker']}{query_type.lower()}/{query}",
            json={"query": query, "type": query_type.lower()},
        ).json()
    except:
        self.rag_response = requests.get(
            f"{rag_response_path['local']}{query_type.lower()}/{query}",
            json={"query": query, "type": query_type.lower()},
        ).json()
    ordered_set = self._order_results()
    self.rag_response["initial_response"] = ordered_set

    return self.rag_response

fetch_structured_query(query_type, query)

Description: Fetch the response for a structured query from the LLM service as a JSON

Source code in frontend/ui_utils.py 
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
def fetch_structured_query(self, query_type: str, query: str):
    """
    Description: Fetch the response for a structured query from the LLM service as a JSON
    """
    structured_response_path = self.paths["structured_query"]
    try:
        self.structured_query_response = requests.get(
            f"{structured_response_path['docker']}{query}",
            json={"query": query},
        ).json()
    except (requests.exceptions.RequestException, json.JSONDecodeError) as e:
        # Print the error for debugging purposes
        print(f"Error occurred: {e}")
        # Set structured_query_response to None on error
        self.structured_query_response = None
    try:
        self.structured_query_response = requests.get(
            f"{structured_response_path['local']}{query}",
            json={"query": query},
        ).json()
    except Exception as e:
        # Print the error for debugging purposes
        print(f"Error occurred while fetching from local endpoint: {e}")
        # Set structured_query_response to None if the local request also fails
        self.structured_query_response = None

    return self.structured_query_response

load_paths()

Description: Load paths from paths.json

Source code in frontend/ui_utils.py 
93
94
95
96
97
98
def load_paths(self):
    """
    Description: Load paths from paths.json
    """
    with open("paths.json", "r") as file:
        return json.load(file)

parse_and_update_response(metadata)

Description: Parse the response from the RAG and LLM services and update the metadata based on the response. Decide which order to apply them - self.apply_llm_before_rag == False - Metadata is filtered based on the rag response first and then by the Query parsing LLM - self.apply_llm_before_rag == False - Metadata is filtered based by the Query parsing LLM first and the rag response second - in case structured_query == true, take results are applying data filters.

Source code in frontend/ui_utils.py 
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
def parse_and_update_response(self, metadata: pd.DataFrame):
    """
     Description: Parse the response from the RAG and LLM services and update the metadata based on the response.
     Decide which order to apply them
     -  self.apply_llm_before_rag == False
         - Metadata is filtered based on the rag response first and then by the Query parsing LLM
    -  self.apply_llm_before_rag == False
         - Metadata is filtered based by the Query parsing LLM first and the rag response second
    - in case structured_query == true, take results are applying data filters.
    """
    if self.apply_llm_before_rag is None or self.llm_response is None:
        print("No LLM filter.")
        # print(self.rag_response, flush=True)
        filtered_metadata = self._no_filter(metadata)

        # print(filtered_metadata)
        # if no llm response is required, return the initial response
        return filtered_metadata

    elif (
        self.rag_response is not None and self.llm_response is not None
    ) and not config["structured_query"]:
        if not self.apply_llm_before_rag:
            filtered_metadata, llm_parser = self._rag_before_llm(metadata)

            if self.query_type.lower() == "dataset":
                llm_parser.get_attributes_from_response()
                return llm_parser.update_subset_cols(filtered_metadata)

        elif self.apply_llm_before_rag:
            filtered_metadata = self._filter_before_rag(metadata)
            return filtered_metadata

    elif (
        self.rag_response is not None and self.structured_query_response is not None
    ):
        col_name = [
            "status",
            "NumberOfClasses",
            "NumberOfFeatures",
            "NumberOfInstances",
        ]
        # print(self.structured_query_response)  # Only for debugging. Comment later.
        if self.structured_query_response[0] is not None and isinstance(
            self.structured_query_response[1], dict
        ):
            # Safely attempt to access the "filter" key in the first element

            self._structured_query_on_success(metadata)

        else:
            filtered_metadata = self._structured_query_on_fail(metadata)
            # print("Showing only rag response")
        return filtered_metadata[["did", "name", *col_name]]

UILoader

Description : Create the chat interface

Source code in frontend/ui_utils.py 
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
class UILoader:
    """
    Description : Create the chat interface
    """

    def __init__(self, config_path):
        with open(config_path, "r") as file:
            # Load config
            self.config = json.load(file)
        # Paths and display information

        # Load metadata chroma database for structured query
        self.collec = load_chroma_metadata()

        # Metadata paths
        self.data_metadata_path = (
            Path(config["data_dir"]) / "all_dataset_description.csv"
        )
        self.flow_metadata_path = Path(config["data_dir"]) / "all_flow_description.csv"

        # Read metadata
        self.data_metadata = pd.read_csv(self.data_metadata_path)
        self.flow_metadata = pd.read_csv(self.flow_metadata_path)

        # defaults
        self.query_type = "Dataset"
        self.llm_filter = False
        self.paths = self.load_paths()
        self.info = """
        <p style='text-align: center; color: white;'>Machine learning research should be easily accessible and reusable. <a href = "https://openml.org/">OpenML</a> is an open platform for sharing datasets, algorithms, and experiments - to learn how to learn better, together. </p>
        """
        self.logo = "images/favicon.ico"
        self.chatbot_display = "How do I do X using OpenML? / Find me a dataset about Y"

        if "messages" not in st.session_state:
            st.session_state.messages = []

    # container for company description and logo
    def _generate_logo_header(
        self,
    ):

        col1, col2 = st.columns([1, 4])
        with col1:
            st.image(self.logo, width=100)
        with col2:
            st.markdown(
                self.info,
                unsafe_allow_html=True,
            )

    def generate_complete_ui(self):

        self._generate_logo_header()
        chat_container = st.container()
        # self.disclaimer_dialog()
        with chat_container:
            with st.form(key="chat_form"):
                user_input = st.text_input(
                    label="Query", placeholder=self.chatbot_display
                )
                query_type = st.selectbox(
                    "Select Query Type",
                    ["General Query", "Dataset", "Flow"],
                    help="Are you looking for a dataset or a flow or just have a general query?",
                )
                ai_filter = st.toggle(
                    "Use AI powered filtering",
                    value=True,
                    help="Uses an AI model to identify what columns might be useful to you.",
                )
                st.form_submit_button(label="Search")

            self.create_chat_interface(user_input=None)
            if user_input:
                self.create_chat_interface(
                    user_input, query_type=query_type, ai_filter=ai_filter
                )

    def create_chat_interface(self, user_input, query_type=None, ai_filter=False):
        """
        Description: Create the chat interface and display the chat history and results. Show the user input and the response from the OpenML Agent.

        """
        self.query_type = query_type
        self.ai_filter = ai_filter

        if user_input is None:
            with st.chat_message(name="ai"):
                st.write("OpenML Agent: ", "Hello! How can I help you today?")
                st.write(
                    ":warning: Note that results are powered by local LLM models and may not be accurate. Please refer to the official OpenML website for accurate information."
                )

        # Handle user input
        if user_input:
            self._handle_user_input(user_input, query_type)

    def _handle_user_input(self, user_input, query_type):
        st.session_state.messages.append({"role": "user", "content": user_input})
        with st.spinner("Waiting for results..."):
            results = self.process_query_chat(user_input)

        if not self.query_type == "General Query":
            st.session_state.messages.append(
                    {"role": "OpenML Agent", "content": results}
                )
        else:
            self._stream_results(results)

            # reverse messages to show the latest message at the top
        reversed_messages = self._reverse_session_history()

            # Display chat history
        self._display_chat_history(query_type, reversed_messages)
        self.create_download_button()

    def _display_chat_history(self, query_type, reversed_messages):
        for message in reversed_messages:
            if query_type == "General Query":
                pass
            if message["role"] == "user":
                with st.chat_message(name="user"):
                    self.display_results(message["content"], "user")
            else:
                with st.chat_message(name="ai"):
                    self.display_results(message["content"], "ai")

    def _reverse_session_history(self):
        reversed_messages = []
        for index in range(0, len(st.session_state.messages), 2):
            reversed_messages.insert(0, st.session_state.messages[index])
            reversed_messages.insert(1, st.session_state.messages[index + 1])
        return reversed_messages

    def _stream_results(self, results):
        with st.spinner("Fetching results..."):
            with requests.get(results, stream=True) as r:
                resp_contain = st.empty()
                streamed_response = ""
                for chunk in r.iter_content(chunk_size=1024):
                    if chunk:
                        streamed_response += chunk.decode("utf-8")
                        resp_contain.markdown(streamed_response)
                resp_contain.empty()
            st.session_state.messages.append(
                {"role": "OpenML Agent", "content": streamed_response}
            )

    @st.experimental_fragment()
    def create_download_button(self):
        data = "\n".join(
            [str(message["content"]) for message in st.session_state.messages]
        )
        st.download_button(
            label="Download chat history",
            data=data,
            file_name="chat_history.txt",
        )

    def display_results(self, initial_response, role):
        """
        Description: Display the results in a DataFrame
        """
        try:
            st.dataframe(initial_response)
        except:
            st.write(initial_response)

    # Function to handle query processing
    def process_query_chat(self, query):
        """
        Description: Process the query and return the results based on the query type and the LLM filter.

        """
        apply_llm_before_rag = None if not self.llm_filter else False
        response_parser = ResponseParser(
            self.query_type, apply_llm_before_rag=apply_llm_before_rag
        )

        if self.query_type == "Dataset" or self.query_type == "Flow":
            if not self.ai_filter:
                response_parser.fetch_rag_response(self.query_type, query)
                return response_parser.parse_and_update_response(self.data_metadata)
            else:
                # get structured query
                self._display_structured_query_results(query, response_parser)

            results = response_parser.parse_and_update_response(self.data_metadata)
            return results

        elif self.query_type == "General Query":
            # Return documentation response path
            return self.paths["documentation_query"]["local"] + query

    def _display_structured_query_results(self, query, response_parser):
        response_parser.fetch_structured_query(self.query_type, query)
        try:
            # get rag response
            # using original query instead of extracted topics.
            response_parser.fetch_rag_response(
                self.query_type,
                response_parser.structured_query_response[0]["query"],
            )

            if response_parser.structured_query_response:
                st.write(
                    "Detected Filter(s): ",
                    json.dumps(
                        response_parser.structured_query_response[0].get("filter", None)
                    ),
                )
            else:
                st.write("Detected Filter(s): ", None)
            if response_parser.structured_query_response[1].get("filter"):
                with st.spinner("Applying LLM Detected Filter(s)..."):
                    response_parser.database_filter(
                        response_parser.structured_query_response[1]["filter"],
                        collec,
                    )
        except:
            # fallback to RAG response
            response_parser.fetch_rag_response(self.query_type, query)

    def load_paths(self):
        """
        Description: Load paths from paths.json
        """
        with open("paths.json", "r") as file:
            return json.load(file)

create_chat_interface(user_input, query_type=None, ai_filter=False)

Description: Create the chat interface and display the chat history and results. Show the user input and the response from the OpenML Agent.

Source code in frontend/ui_utils.py 
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
def create_chat_interface(self, user_input, query_type=None, ai_filter=False):
    """
    Description: Create the chat interface and display the chat history and results. Show the user input and the response from the OpenML Agent.

    """
    self.query_type = query_type
    self.ai_filter = ai_filter

    if user_input is None:
        with st.chat_message(name="ai"):
            st.write("OpenML Agent: ", "Hello! How can I help you today?")
            st.write(
                ":warning: Note that results are powered by local LLM models and may not be accurate. Please refer to the official OpenML website for accurate information."
            )

    # Handle user input
    if user_input:
        self._handle_user_input(user_input, query_type)

display_results(initial_response, role)

Description: Display the results in a DataFrame

Source code in frontend/ui_utils.py 
477
478
479
480
481
482
483
484
def display_results(self, initial_response, role):
    """
    Description: Display the results in a DataFrame
    """
    try:
        st.dataframe(initial_response)
    except:
        st.write(initial_response)

load_paths()

Description: Load paths from paths.json

Source code in frontend/ui_utils.py 
541
542
543
544
545
546
def load_paths(self):
    """
    Description: Load paths from paths.json
    """
    with open("paths.json", "r") as file:
        return json.load(file)

process_query_chat(query)

Description: Process the query and return the results based on the query type and the LLM filter.

Source code in frontend/ui_utils.py 
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
def process_query_chat(self, query):
    """
    Description: Process the query and return the results based on the query type and the LLM filter.

    """
    apply_llm_before_rag = None if not self.llm_filter else False
    response_parser = ResponseParser(
        self.query_type, apply_llm_before_rag=apply_llm_before_rag
    )

    if self.query_type == "Dataset" or self.query_type == "Flow":
        if not self.ai_filter:
            response_parser.fetch_rag_response(self.query_type, query)
            return response_parser.parse_and_update_response(self.data_metadata)
        else:
            # get structured query
            self._display_structured_query_results(query, response_parser)

        results = response_parser.parse_and_update_response(self.data_metadata)
        return results

    elif self.query_type == "General Query":
        # Return documentation response path
        return self.paths["documentation_query"]["local"] + query