CSS update + IPOS
Browse files- README.md +1 -1
- app.py +33 -220
- climateqa/engine/retriever.py +4 -3
- climateqa/sample_questions.py +20 -1
- requirements.txt +1 -1
- style.css +39 -0
README.md
CHANGED
@@ -4,7 +4,7 @@ emoji: π
|
|
4 |
colorFrom: blue
|
5 |
colorTo: red
|
6 |
sdk: gradio
|
7 |
-
sdk_version: 4.
|
8 |
app_file: app.py
|
9 |
fullWidth: true
|
10 |
pinned: false
|
|
|
4 |
colorFrom: blue
|
5 |
colorTo: red
|
6 |
sdk: gradio
|
7 |
+
sdk_version: 4.19.1
|
8 |
app_file: app.py
|
9 |
fullWidth: true
|
10 |
pinned: false
|
app.py
CHANGED
@@ -10,6 +10,8 @@ import time
|
|
10 |
import re
|
11 |
import json
|
12 |
|
|
|
|
|
13 |
from io import BytesIO
|
14 |
import base64
|
15 |
|
@@ -108,28 +110,13 @@ def serialize_docs(docs):
|
|
108 |
return new_docs
|
109 |
|
110 |
|
111 |
-
# import asyncio
|
112 |
-
# from typing import Any, Dict, List
|
113 |
-
# from langchain.callbacks.base import AsyncCallbackHandler, BaseCallbackHandler
|
114 |
-
|
115 |
-
# class MyCustomAsyncHandler(AsyncCallbackHandler):
|
116 |
-
# """Async callback handler that can be used to handle callbacks from langchain."""
|
117 |
-
|
118 |
-
# async def on_chain_start(
|
119 |
-
# self, serialized: Dict[str, Any], inputs: Dict[str, Any], **kwargs: Any
|
120 |
-
# ) -> Any:
|
121 |
-
# """Run when chain starts running."""
|
122 |
-
# print("zzzz....")
|
123 |
-
# await asyncio.sleep(3)
|
124 |
-
# print(f"on_chain_start {serialized['name']}")
|
125 |
-
# # raise gr.Error("ClimateQ&A Error: Timeout, try another question and if the error remains, you can contact us :)")
|
126 |
-
|
127 |
-
|
128 |
|
129 |
async def chat(query,history,audience,sources,reports):
|
130 |
"""taking a query and a message history, use a pipeline (reformulation, retriever, answering) to yield a tuple of:
|
131 |
(messages in gradio format, messages in langchain format, source documents)"""
|
132 |
|
|
|
|
|
133 |
if audience == "Children":
|
134 |
audience_prompt = audience_prompts["children"]
|
135 |
elif audience == "General public":
|
@@ -189,8 +176,10 @@ async def chat(query,history,audience,sources,reports):
|
|
189 |
|
190 |
elif op['path'] == path_answer: # final answer
|
191 |
new_token = op['value'] # str
|
192 |
-
time.sleep(0.01)
|
193 |
-
|
|
|
|
|
194 |
answer_yet = parse_output_llm_with_sources(answer_yet)
|
195 |
history[-1] = (query,answer_yet)
|
196 |
|
@@ -324,16 +313,12 @@ def make_html_source(source,i):
|
|
324 |
else:
|
325 |
break
|
326 |
toc_levels = " > ".join(toc_levels)
|
327 |
-
print(toc_levels)
|
328 |
|
329 |
if len(toc_levels) > 0:
|
330 |
name = f"<b>{toc_levels}</b><br/>{meta['name']}"
|
331 |
else:
|
332 |
name = meta['name']
|
333 |
|
334 |
-
print(name)
|
335 |
-
|
336 |
-
|
337 |
if meta["chunk_type"] == "text":
|
338 |
|
339 |
card = f"""
|
@@ -404,12 +389,7 @@ def save_feedback(feed: str, user_id):
|
|
404 |
def log_on_azure(file, logs, share_client):
|
405 |
logs = json.dumps(logs)
|
406 |
file_client = share_client.get_file_client(file)
|
407 |
-
print("Uploading logs to Azure Blob Storage")
|
408 |
-
print("----------------------------------")
|
409 |
-
print("")
|
410 |
-
print(logs)
|
411 |
file_client.upload_file(logs)
|
412 |
-
print("Logs uploaded to Azure Blob Storage")
|
413 |
|
414 |
|
415 |
# --------------------------------------------------------------------
|
@@ -459,7 +439,8 @@ with gr.Blocks(title="Climate Q&A", css="style.css", theme=theme,elem_id = "main
|
|
459 |
|
460 |
|
461 |
with gr.Row(elem_id = "input-message"):
|
462 |
-
textbox=gr.Textbox(placeholder="Ask me anything here!",show_label=False,scale=
|
|
|
463 |
|
464 |
|
465 |
with gr.Column(scale=1, variant="panel",elem_id = "right-panel"):
|
@@ -497,13 +478,14 @@ with gr.Blocks(title="Climate Q&A", css="style.css", theme=theme,elem_id = "main
|
|
497 |
sources_textbox = gr.HTML(show_label=False, elem_id="sources-textbox")
|
498 |
docs_textbox = gr.State("")
|
499 |
|
|
|
500 |
with gr.Tab("Configuration",elem_id = "tab-config",id = 2):
|
501 |
|
502 |
gr.Markdown("Reminder: You can talk in any language, ClimateQ&A is multi-lingual!")
|
503 |
|
504 |
|
505 |
dropdown_sources = gr.CheckboxGroup(
|
506 |
-
["IPCC", "IPBES"],
|
507 |
label="Select source",
|
508 |
value=["IPCC"],
|
509 |
interactive=True,
|
@@ -539,202 +521,15 @@ with gr.Blocks(title="Climate Q&A", css="style.css", theme=theme,elem_id = "main
|
|
539 |
with gr.Tab("Figures",elem_id = "tab-images",elem_classes = "max-height other-tabs"):
|
540 |
gallery_component = gr.Gallery()
|
541 |
|
542 |
-
with gr.Tab("About
|
543 |
with gr.Row():
|
544 |
with gr.Column(scale=1):
|
545 |
-
gr.Markdown(
|
546 |
-
"""
|
547 |
-
<p><b>Climate change and environmental disruptions have become some of the most pressing challenges facing our planet today</b>. As global temperatures rise and ecosystems suffer, it is essential for individuals to understand the gravity of the situation in order to make informed decisions and advocate for appropriate policy changes.</p>
|
548 |
-
<p>However, comprehending the vast and complex scientific information can be daunting, as the scientific consensus references, such as <b>the Intergovernmental Panel on Climate Change (IPCC) reports, span thousands of pages</b>. To bridge this gap and make climate science more accessible, we introduce <b>ClimateQ&A as a tool to distill expert-level knowledge into easily digestible insights about climate science.</b></p>
|
549 |
-
<div class="tip-box">
|
550 |
-
<div class="tip-box-title">
|
551 |
-
<span class="light-bulb" role="img" aria-label="Light Bulb">π‘</span>
|
552 |
-
How does ClimateQ&A work?
|
553 |
-
</div>
|
554 |
-
ClimateQ&A harnesses modern OCR techniques to parse and preprocess IPCC reports. By leveraging state-of-the-art question-answering algorithms, <i>ClimateQ&A is able to sift through the extensive collection of climate scientific reports and identify relevant passages in response to user inquiries</i>. Furthermore, the integration of the ChatGPT API allows ClimateQ&A to present complex data in a user-friendly manner, summarizing key points and facilitating communication of climate science to a wider audience.
|
555 |
-
</div>
|
556 |
-
"""
|
557 |
-
)
|
558 |
-
|
559 |
-
with gr.Column(scale=1):
|
560 |
-
gr.Markdown("![](https://i.postimg.cc/fLvsvMzM/Untitled-design-5.png)")
|
561 |
-
gr.Markdown("*Source : IPCC AR6 - Synthesis Report of the IPCC 6th assessment report (AR6)*")
|
562 |
-
|
563 |
-
gr.Markdown("## How to use ClimateQ&A")
|
564 |
-
with gr.Row():
|
565 |
-
with gr.Column(scale=1):
|
566 |
-
gr.Markdown(
|
567 |
-
"""
|
568 |
-
### Getting started
|
569 |
-
- In the chatbot section, simply type your climate-related question, and ClimateQ&A will provide an answer with references to relevant IPCC reports.
|
570 |
-
- ClimateQ&A retrieves specific passages from the IPCC reports to help answer your question accurately.
|
571 |
-
- Source information, including page numbers and passages, is displayed on the right side of the screen for easy verification.
|
572 |
-
- Feel free to ask follow-up questions within the chatbot for a more in-depth understanding.
|
573 |
-
- You can ask question in any language, ClimateQ&A is multi-lingual !
|
574 |
-
- ClimateQ&A integrates multiple sources (IPCC and IPBES, β¦ ) to cover various aspects of environmental science, such as climate change and biodiversity. See all sources used below.
|
575 |
-
"""
|
576 |
-
)
|
577 |
-
with gr.Column(scale=1):
|
578 |
-
gr.Markdown(
|
579 |
-
"""
|
580 |
-
### Limitations
|
581 |
-
<div class="warning-box">
|
582 |
-
<ul>
|
583 |
-
<li>Please note that, like any AI, the model may occasionally generate an inaccurate or imprecise answer. Always refer to the provided sources to verify the validity of the information given. If you find any issues with the response, kindly provide feedback to help improve the system.</li>
|
584 |
-
<li>ClimateQ&A is specifically designed for climate-related inquiries. If you ask a non-environmental question, the chatbot will politely remind you that its focus is on climate and environmental issues.</li>
|
585 |
-
</div>
|
586 |
-
"""
|
587 |
-
)
|
588 |
-
|
589 |
-
|
590 |
-
with gr.Tab("Contact, feedback and feature requests",elem_classes = "max-height other-tabs"):
|
591 |
-
gr.Markdown(
|
592 |
-
"""
|
593 |
-
|
594 |
-
For any question or press request, contact ThΓ©o Alves Da Costa at <b>theo.alvesdacosta@ekimetrics.com</b>
|
595 |
-
|
596 |
-
- ClimateQ&A welcomes community contributions. To participate, head over to the Community Tab and create a "New Discussion" to ask questions and share your insights.
|
597 |
-
- Provide feedback through email, letting us know which insights you found accurate, useful, or not. Your input will help us improve the platform.
|
598 |
-
- Only a few sources (see below) are integrated (all IPCC, IPBES), if you are a climate science researcher and net to sift through another report, please let us know.
|
599 |
-
|
600 |
-
*This tool has been developed by the R&D lab at **Ekimetrics** (Jean Lelong, Nina Achache, Gabriel Olympie, Nicolas Chesneau, Natalia De la Calzada, ThΓ©o Alves Da Costa)*
|
601 |
-
"""
|
602 |
-
)
|
603 |
-
# with gr.Row():
|
604 |
-
# with gr.Column(scale=1):
|
605 |
-
# gr.Markdown("### Feedbacks")
|
606 |
-
# feedback = gr.Textbox(label="Write your feedback here")
|
607 |
-
# feedback_output = gr.Textbox(label="Submit status")
|
608 |
-
# feedback_save = gr.Button(value="submit feedback")
|
609 |
-
# feedback_save.click(
|
610 |
-
# save_feedback,
|
611 |
-
# inputs=[feedback, user_id_state],
|
612 |
-
# outputs=feedback_output,
|
613 |
-
# )
|
614 |
-
# gr.Markdown(
|
615 |
-
# "If you need us to ask another climate science report or ask any question, contact us at <b>theo.alvesdacosta@ekimetrics.com</b>"
|
616 |
-
# )
|
617 |
-
|
618 |
-
# with gr.Column(scale=1):
|
619 |
-
# gr.Markdown("### OpenAI API")
|
620 |
-
# gr.Markdown(
|
621 |
-
# "To make climate science accessible to a wider audience, we have opened our own OpenAI API key with a monthly cap of $1000. If you already have an API key, please use it to help conserve bandwidth for others."
|
622 |
-
# )
|
623 |
-
# openai_api_key_textbox = gr.Textbox(
|
624 |
-
# placeholder="Paste your OpenAI API key (sk-...) and hit Enter",
|
625 |
-
# show_label=False,
|
626 |
-
# lines=1,
|
627 |
-
# type="password",
|
628 |
-
# )
|
629 |
-
# openai_api_key_textbox.change(set_openai_api_key, inputs=[openai_api_key_textbox])
|
630 |
-
# openai_api_key_textbox.submit(set_openai_api_key, inputs=[openai_api_key_textbox])
|
631 |
-
|
632 |
-
with gr.Tab("Sources",elem_classes = "max-height other-tabs"):
|
633 |
-
gr.Markdown("""
|
634 |
-
| Source | Report | URL | Number of pages | Release date |
|
635 |
-
| --- | --- | --- | --- | --- |
|
636 |
-
IPCC | Summary for Policymakers. In: Climate Change 2021: The Physical Science Basis. Contribution of the WGI to the AR6 of the IPCC. | https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_SPM.pdf | 32 | 2021
|
637 |
-
IPCC | Full Report. In: Climate Change 2021: The Physical Science Basis. Contribution of the WGI to the AR6 of the IPCC. | https://report.ipcc.ch/ar6/wg1/IPCC_AR6_WGI_FullReport.pdf | 2409 | 2021
|
638 |
-
IPCC | Technical Summary. In: Climate Change 2021: The Physical Science Basis. Contribution of the WGI to the AR6 of the IPCC. | https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_TS.pdf | 112 | 2021
|
639 |
-
IPCC | Summary for Policymakers. In: Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of the WGII to the AR6 of the IPCC. | https://www.ipcc.ch/report/ar6/wg2/downloads/report/IPCC_AR6_WGII_SummaryForPolicymakers.pdf | 34 | 2022
|
640 |
-
IPCC | Technical Summary. In: Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of the WGII to the AR6 of the IPCC. | https://www.ipcc.ch/report/ar6/wg2/downloads/report/IPCC_AR6_WGII_TechnicalSummary.pdf | 84 | 2022
|
641 |
-
IPCC | Full Report. In: Climate Change 2022: Impacts, Adaptation and Vulnerability. Contribution of the WGII to the AR6 of the IPCC. | https://report.ipcc.ch/ar6/wg2/IPCC_AR6_WGII_FullReport.pdf | 3068 | 2022
|
642 |
-
IPCC | Summary for Policymakers. In: Climate Change 2022: Mitigation of Climate Change. Contribution of the WGIII to the AR6 of the IPCC. | https://www.ipcc.ch/report/ar6/wg3/downloads/report/IPCC_AR6_WGIII_SummaryForPolicymakers.pdf | 50 | 2022
|
643 |
-
IPCC | Technical Summary. In: Climate Change 2022: Mitigation of Climate Change. Contribution of the WGIII to the AR6 of the IPCC. | https://www.ipcc.ch/report/ar6/wg3/downloads/report/IPCC_AR6_WGIII_TechnicalSummary.pdf | 102 | 2022
|
644 |
-
IPCC | Full Report. In: Climate Change 2022: Mitigation of Climate Change. Contribution of the WGIII to the AR6 of the IPCC. | https://www.ipcc.ch/report/ar6/wg3/downloads/report/IPCC_AR6_WGIII_FullReport.pdf | 2258 | 2022
|
645 |
-
IPCC | Summary for Policymakers. In: Global Warming of 1.5ΓΒ°C. An IPCC Special Report on the impacts of global warming of 1.5ΓΒ°C above pre-industrial levels and related global greenhouse gas emission pathways, in the context of strengthening the global response to the threat of climate change, sustainable development, and efforts to eradicate poverty. | https://www.ipcc.ch/site/assets/uploads/sites/2/2022/06/SPM_version_report_LR.pdf | 24 | 2018
|
646 |
-
IPCC | Summary for Policymakers. In: Climate Change and Land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems. | https://www.ipcc.ch/site/assets/uploads/sites/4/2022/11/SRCCL_SPM.pdf | 36 | 2019
|
647 |
-
IPCC | Summary for Policymakers. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/01_SROCC_SPM_FINAL.pdf | 36 | 2019
|
648 |
-
IPCC | Technical Summary. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/02_SROCC_TS_FINAL.pdf | 34 | 2019
|
649 |
-
IPCC | Chapter 1 - Framing and Context of the Report. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/03_SROCC_Ch01_FINAL.pdf | 60 | 2019
|
650 |
-
IPCC | Chapter 2 - High Mountain Areas. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/04_SROCC_Ch02_FINAL.pdf | 72 | 2019
|
651 |
-
IPCC | Chapter 3 - Polar Regions. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/05_SROCC_Ch03_FINAL.pdf | 118 | 2019
|
652 |
-
IPCC | Chapter 4 - Sea Level Rise and Implications for Low-Lying Islands, Coasts and Communities. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/06_SROCC_Ch04_FINAL.pdf | 126 | 2019
|
653 |
-
IPCC | Chapter 5 - Changing Ocean, Marine Ecosystems, and Dependent Communities. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/07_SROCC_Ch05_FINAL.pdf | 142 | 2019
|
654 |
-
IPCC | Chapter 6 - Extremes, Abrupt Changes and Managing Risk. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/08_SROCC_Ch06_FINAL.pdf | 68 | 2019
|
655 |
-
IPCC | Cross-Chapter Box 9: Integrative Cross-Chapter Box on Low-Lying Islands and Coasts. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2019/11/11_SROCC_CCB9-LLIC_FINAL.pdf | 18 | 2019
|
656 |
-
IPCC | Annex I: Glossary [Weyer, N.M. (ed.)]. In: IPCC Special Report on the Ocean and Cryosphere in a Changing Climate. | https://www.ipcc.ch/site/assets/uploads/sites/3/2022/03/10_SROCC_AnnexI-Glossary_FINAL.pdf | 28 | 2019
|
657 |
-
IPBES | Full Report. Global assessment report on biodiversity and ecosystem services of the IPBES. | https://zenodo.org/record/6417333/files/202206_IPBES%20GLOBAL%20REPORT_FULL_DIGITAL_MARCH%202022.pdf | 1148 | 2019
|
658 |
-
IPBES | Summary for Policymakers. Global assessment report on biodiversity and ecosystem services of the IPBES (Version 1). | https://zenodo.org/record/3553579/files/ipbes_global_assessment_report_summary_for_policymakers.pdf | 60 | 2019
|
659 |
-
IPBES | Full Report. Thematic assessment of the sustainable use of wild species of the IPBES. | https://zenodo.org/record/7755805/files/IPBES_ASSESSMENT_SUWS_FULL_REPORT.pdf | 1008 | 2022
|
660 |
-
IPBES | Summary for Policymakers. Summary for policymakers of the thematic assessment of the sustainable use of wild species of the IPBES. | https://zenodo.org/record/7411847/files/EN_SPM_SUSTAINABLE%20USE%20OF%20WILD%20SPECIES.pdf | 44 | 2022
|
661 |
-
IPBES | Full Report. Regional Assessment Report on Biodiversity and Ecosystem Services for Africa. | https://zenodo.org/record/3236178/files/ipbes_assessment_report_africa_EN.pdf | 494 | 2018
|
662 |
-
IPBES | Summary for Policymakers. Regional Assessment Report on Biodiversity and Ecosystem Services for Africa. | https://zenodo.org/record/3236189/files/ipbes_assessment_spm_africa_EN.pdf | 52 | 2018
|
663 |
-
IPBES | Full Report. Regional Assessment Report on Biodiversity and Ecosystem Services for the Americas. | https://zenodo.org/record/3236253/files/ipbes_assessment_report_americas_EN.pdf | 660 | 2018
|
664 |
-
IPBES | Summary for Policymakers. Regional Assessment Report on Biodiversity and Ecosystem Services for the Americas. | https://zenodo.org/record/3236292/files/ipbes_assessment_spm_americas_EN.pdf | 44 | 2018
|
665 |
-
IPBES | Full Report. Regional Assessment Report on Biodiversity and Ecosystem Services for Asia and the Pacific. | https://zenodo.org/record/3237374/files/ipbes_assessment_report_ap_EN.pdf | 616 | 2018
|
666 |
-
IPBES | Summary for Policymakers. Regional Assessment Report on Biodiversity and Ecosystem Services for Asia and the Pacific. | https://zenodo.org/record/3237383/files/ipbes_assessment_spm_ap_EN.pdf | 44 | 2018
|
667 |
-
IPBES | Full Report. Regional Assessment Report on Biodiversity and Ecosystem Services for Europe and Central Asia. | https://zenodo.org/record/3237429/files/ipbes_assessment_report_eca_EN.pdf | 894 | 2018
|
668 |
-
IPBES | Summary for Policymakers. Regional Assessment Report on Biodiversity and Ecosystem Services for Europe and Central Asia. | https://zenodo.org/record/3237468/files/ipbes_assessment_spm_eca_EN.pdf | 52 | 2018
|
669 |
-
IPBES | Full Report. Assessment Report on Land Degradation and Restoration. | https://zenodo.org/record/3237393/files/ipbes_assessment_report_ldra_EN.pdf | 748 | 2018
|
670 |
-
IPBES | IPBES Invasive Alien Species Assessment: Summary for Policymakers & 6 chapters | https://zenodo.org/records/10127924/files/Summary%20for%20policymakers_IPBES%20IAS%20Assessment.pdf | 56 + 1198 | 2023
|
671 |
-
""")
|
672 |
-
|
673 |
-
with gr.Tab("Carbon Footprint",elem_classes = "max-height other-tabs"):
|
674 |
-
gr.Markdown("""
|
675 |
-
|
676 |
-
Carbon emissions were measured during the development and inference process using CodeCarbon [https://github.com/mlco2/codecarbon](https://github.com/mlco2/codecarbon)
|
677 |
-
|
678 |
-
| Phase | Description | Emissions | Source |
|
679 |
-
| --- | --- | --- | --- |
|
680 |
-
| Development | OCR and parsing all pdf documents with AI | 28gCO2e | CodeCarbon |
|
681 |
-
| Development | Question Answering development | 114gCO2e | CodeCarbon |
|
682 |
-
| Inference | Question Answering | ~0.102gCO2e / call | CodeCarbon |
|
683 |
-
| Inference | API call to turbo-GPT | ~0.38gCO2e / call | https://medium.com/@chrispointon/the-carbon-footprint-of-chatgpt-e1bc14e4cc2a |
|
684 |
-
|
685 |
-
Carbon Emissions are **relatively low but not negligible** compared to other usages: one question asked to ClimateQ&A is around 0.482gCO2e - equivalent to 2.2m by car (https://datagir.ademe.fr/apps/impact-co2/)
|
686 |
-
Or around 2 to 4 times more than a typical Google search.
|
687 |
-
"""
|
688 |
-
)
|
689 |
-
|
690 |
-
with gr.Tab("Changelog",elem_classes = "max-height other-tabs"):
|
691 |
-
gr.Markdown("""
|
692 |
-
|
693 |
-
##### Upcoming features
|
694 |
-
- Figures retrieval and multimodal system
|
695 |
-
- Conversational chat
|
696 |
-
- Intent routing
|
697 |
-
- Local environment setup
|
698 |
-
|
699 |
-
##### v1.3.0-alpha - *2024-01-31*
|
700 |
-
- Completely redesign chunking system mechanism at the parsing phase (hierarchical aggregations, better control to get 256 tokens)
|
701 |
-
- Switched to Pinecone serverless for performance improvements
|
702 |
-
- Add TOC section for each source in the citations panel as breadcrumbs
|
703 |
-
- *EXPERIMENTAL* - Figure retrieval system within the answer (caption pictures with GPT4 Vision and indexing the caption in the retrieval system)
|
704 |
-
|
705 |
-
##### v1.2.1 - *2024-01-16*
|
706 |
-
- *BUG* - corrected asynchronous bug failing the chatbot frequently
|
707 |
-
|
708 |
-
##### v1.2.0 - *2023-11-27*
|
709 |
-
- Added new IPBES assessment on Invasive Species (SPM and chapters)
|
710 |
-
- Switched all the codebase to LCEL (Langchain Expression Language)
|
711 |
-
- Added sample questions by category
|
712 |
-
- Switched embeddings from old ``sentence-transformers/multi-qa-mpnet-base-dot-v1`` to ``BAAI/bge-base-en-v1.5``
|
713 |
-
- Report filtering to select directly the report you want to source your answers from
|
714 |
-
- First naive version of a figures retrieval system by looking up the figures in the retrieved pages
|
715 |
-
|
716 |
-
##### v1.1.0 - *2023-10-16*
|
717 |
-
- ClimateQ&A on Hugging Face is finally working again with all the new features !
|
718 |
-
- Switched all python code to langchain codebase for cleaner code, easier maintenance and future features
|
719 |
-
- Updated GPT model to August version
|
720 |
-
- Added streaming response to improve UX
|
721 |
-
- Created a custom Retriever chain to avoid calling the LLM if there is no documents retrieved
|
722 |
-
- Use of HuggingFace embed on https://climateqa.com to avoid demultiplying deployments
|
723 |
-
|
724 |
-
##### v1.0.0 - *2023-05-11*
|
725 |
-
- First version of clean interface on https://climateqa.com
|
726 |
-
- Add children mode on https://climateqa.com
|
727 |
-
- Add follow-up questions https://climateqa.com
|
728 |
-
"""
|
729 |
-
)
|
730 |
-
|
731 |
-
|
732 |
|
733 |
|
734 |
def start_chat(query,history):
|
735 |
-
history = history + [(query,
|
736 |
history = [tuple(x) for x in history]
|
737 |
-
print(history)
|
738 |
return (gr.update(interactive = False),gr.update(selected=1),history)
|
739 |
|
740 |
def finish_chat():
|
@@ -784,6 +579,24 @@ Or around 2 to 4 times more than a typical Google search.
|
|
784 |
# )
|
785 |
|
786 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
787 |
demo.queue()
|
788 |
|
789 |
demo.launch()
|
|
|
10 |
import re
|
11 |
import json
|
12 |
|
13 |
+
from gradio_modal import Modal
|
14 |
+
|
15 |
from io import BytesIO
|
16 |
import base64
|
17 |
|
|
|
110 |
return new_docs
|
111 |
|
112 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
113 |
|
114 |
async def chat(query,history,audience,sources,reports):
|
115 |
"""taking a query and a message history, use a pipeline (reformulation, retriever, answering) to yield a tuple of:
|
116 |
(messages in gradio format, messages in langchain format, source documents)"""
|
117 |
|
118 |
+
print(f">> NEW QUESTION : {query}")
|
119 |
+
|
120 |
if audience == "Children":
|
121 |
audience_prompt = audience_prompts["children"]
|
122 |
elif audience == "General public":
|
|
|
176 |
|
177 |
elif op['path'] == path_answer: # final answer
|
178 |
new_token = op['value'] # str
|
179 |
+
# time.sleep(0.01)
|
180 |
+
previous_answer = history[-1][1]
|
181 |
+
previous_answer = previous_answer if previous_answer is not None else ""
|
182 |
+
answer_yet = previous_answer + new_token
|
183 |
answer_yet = parse_output_llm_with_sources(answer_yet)
|
184 |
history[-1] = (query,answer_yet)
|
185 |
|
|
|
313 |
else:
|
314 |
break
|
315 |
toc_levels = " > ".join(toc_levels)
|
|
|
316 |
|
317 |
if len(toc_levels) > 0:
|
318 |
name = f"<b>{toc_levels}</b><br/>{meta['name']}"
|
319 |
else:
|
320 |
name = meta['name']
|
321 |
|
|
|
|
|
|
|
322 |
if meta["chunk_type"] == "text":
|
323 |
|
324 |
card = f"""
|
|
|
389 |
def log_on_azure(file, logs, share_client):
|
390 |
logs = json.dumps(logs)
|
391 |
file_client = share_client.get_file_client(file)
|
|
|
|
|
|
|
|
|
392 |
file_client.upload_file(logs)
|
|
|
393 |
|
394 |
|
395 |
# --------------------------------------------------------------------
|
|
|
439 |
|
440 |
|
441 |
with gr.Row(elem_id = "input-message"):
|
442 |
+
textbox=gr.Textbox(placeholder="Ask me anything here!",show_label=False,scale=7,lines = 1,interactive = True,elem_id="input-textbox")
|
443 |
+
# submit = gr.Button("",elem_id = "submit-button",scale = 1,interactive = True,icon = "https://static-00.iconduck.com/assets.00/settings-icon-2048x2046-cw28eevx.png")
|
444 |
|
445 |
|
446 |
with gr.Column(scale=1, variant="panel",elem_id = "right-panel"):
|
|
|
478 |
sources_textbox = gr.HTML(show_label=False, elem_id="sources-textbox")
|
479 |
docs_textbox = gr.State("")
|
480 |
|
481 |
+
# with Modal(visible = False) as config_modal:
|
482 |
with gr.Tab("Configuration",elem_id = "tab-config",id = 2):
|
483 |
|
484 |
gr.Markdown("Reminder: You can talk in any language, ClimateQ&A is multi-lingual!")
|
485 |
|
486 |
|
487 |
dropdown_sources = gr.CheckboxGroup(
|
488 |
+
["IPCC", "IPBES","IPOS"],
|
489 |
label="Select source",
|
490 |
value=["IPCC"],
|
491 |
interactive=True,
|
|
|
521 |
with gr.Tab("Figures",elem_id = "tab-images",elem_classes = "max-height other-tabs"):
|
522 |
gallery_component = gr.Gallery()
|
523 |
|
524 |
+
with gr.Tab("About",elem_classes = "max-height other-tabs"):
|
525 |
with gr.Row():
|
526 |
with gr.Column(scale=1):
|
527 |
+
gr.Markdown("See more info at [https://climateqa.com](https://climateqa.com/docs/intro/)")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
528 |
|
529 |
|
530 |
def start_chat(query,history):
|
531 |
+
history = history + [(query,None)]
|
532 |
history = [tuple(x) for x in history]
|
|
|
533 |
return (gr.update(interactive = False),gr.update(selected=1),history)
|
534 |
|
535 |
def finish_chat():
|
|
|
579 |
# )
|
580 |
|
581 |
|
582 |
+
# with Modal(visible=True) as first_modal:
|
583 |
+
# gr.Markdown("# Welcome to ClimateQ&A !")
|
584 |
+
|
585 |
+
# gr.Markdown("### Examples")
|
586 |
+
|
587 |
+
# examples = gr.Examples(
|
588 |
+
# ["Yo Γ§a roule","Γ§a boume"],
|
589 |
+
# [examples_hidden],
|
590 |
+
# examples_per_page=8,
|
591 |
+
# run_on_click=False,
|
592 |
+
# elem_id="examples",
|
593 |
+
# api_name="examples",
|
594 |
+
# )
|
595 |
+
|
596 |
+
|
597 |
+
# submit.click(lambda: Modal(visible=True), None, config_modal)
|
598 |
+
|
599 |
+
|
600 |
demo.queue()
|
601 |
|
602 |
demo.launch()
|
climateqa/engine/retriever.py
CHANGED
@@ -13,7 +13,7 @@ from pydantic import Field
|
|
13 |
|
14 |
class ClimateQARetriever(BaseRetriever):
|
15 |
vectorstore:VectorStore
|
16 |
-
sources:list = ["IPCC","IPBES"]
|
17 |
reports:list = []
|
18 |
threshold:float = 0.6
|
19 |
k_summary:int = 3
|
@@ -28,7 +28,7 @@ class ClimateQARetriever(BaseRetriever):
|
|
28 |
|
29 |
# Check if all elements in the list are either IPCC or IPBES
|
30 |
assert isinstance(self.sources,list)
|
31 |
-
assert all([x in ["IPCC","IPBES"] for x in self.sources])
|
32 |
assert self.k_total > self.k_summary, "k_total should be greater than k_summary"
|
33 |
|
34 |
# Prepare base search kwargs
|
@@ -60,7 +60,8 @@ class ClimateQARetriever(BaseRetriever):
|
|
60 |
docs = docs_summaries + docs_full
|
61 |
|
62 |
# Filter if scores are below threshold
|
63 |
-
|
|
|
64 |
|
65 |
# Add score to metadata
|
66 |
results = []
|
|
|
13 |
|
14 |
class ClimateQARetriever(BaseRetriever):
|
15 |
vectorstore:VectorStore
|
16 |
+
sources:list = ["IPCC","IPBES","IPOS"]
|
17 |
reports:list = []
|
18 |
threshold:float = 0.6
|
19 |
k_summary:int = 3
|
|
|
28 |
|
29 |
# Check if all elements in the list are either IPCC or IPBES
|
30 |
assert isinstance(self.sources,list)
|
31 |
+
assert all([x in ["IPCC","IPBES","IPOS"] for x in self.sources])
|
32 |
assert self.k_total > self.k_summary, "k_total should be greater than k_summary"
|
33 |
|
34 |
# Prepare base search kwargs
|
|
|
60 |
docs = docs_summaries + docs_full
|
61 |
|
62 |
# Filter if scores are below threshold
|
63 |
+
docs = [x for x in docs if len(x[0].page_content) > self.min_size]
|
64 |
+
# docs = [x for x in docs if x[1] > self.threshold]
|
65 |
|
66 |
# Add score to metadata
|
67 |
results = []
|
climateqa/sample_questions.py
CHANGED
@@ -80,6 +80,25 @@ QUESTIONS = {
|
|
80 |
"What is the distribution of uncertainty in projected precipitation changes across different time frames ?",
|
81 |
"What are the anticipated changes in the global water cycle by the end of the 21st century under an intermediate emissions scenario ?",
|
82 |
"How are human activities contributing to the spread of invasive alien species?",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
83 |
]
|
84 |
-
|
85 |
}
|
|
|
80 |
"What is the distribution of uncertainty in projected precipitation changes across different time frames ?",
|
81 |
"What are the anticipated changes in the global water cycle by the end of the 21st century under an intermediate emissions scenario ?",
|
82 |
"How are human activities contributing to the spread of invasive alien species?",
|
83 |
+
],
|
84 |
+
"Deep Sea Mining":[
|
85 |
+
"What is the motivation behind mining the deep seabed?",
|
86 |
+
"What are the arguments in favor of and against deep-sea mining?",
|
87 |
+
"Can the global demand for metals and resources be met adequately through land-based sources for the next few decades?",
|
88 |
+
"Is it essential for humanity to exploit the deep ocean for minerals, or should we consider alternative pathways?",
|
89 |
+
"What are the environmental impacts of deep-sea mining in order of significance, indicating the degree or severity of the impacts?",
|
90 |
+
"What specific substances are likely to be released as contaminants through deep-sea mining?",
|
91 |
+
"How might the substances released through deep-sea mining pose a threat to marine life?",
|
92 |
+
"What are the various environmental stressors associated with deep-sea mining, including factors like noise, vibration, and light?",
|
93 |
+
"How would environmental stressors associated with deep-sea mining affect marine ecosystems?",
|
94 |
+
"How will the removal of mineral resources as a result of deep-sea mining impact the living components of the ecosystem?",
|
95 |
+
"How will deep-sea mining influence ocean currents, large-scale circulation, and biogeochemical cycles?",
|
96 |
+
"What are the potential ramifications of deep-sea mining for global ocean processes?",
|
97 |
+
"What are the potential repercussions of deep-sea mining on climate regulation, considering the role of deep-sea ecosystems in climate dynamics?",
|
98 |
+
"How resilient are the targeted habitats of deep-sea ecosystems (nodules, sulfides, crusts) to the impacts of deep-sea mining?",
|
99 |
+
"Apart from the ecological consequences of deep-sea mining, what economic challenges could a state encounter if a deep-seabed mining operation leads to financial losses or third-party liability?",
|
100 |
+
"What legal steps would need to be designed to authorize deep-sea mining?",
|
101 |
+
"Is the current technological infrastructure sufficiently advanced and tested to support the implementation of deep-sea mining operations effectively?",
|
102 |
+
"Provide me with a list of organizations most actively opposing deep-sea mining."
|
103 |
]
|
|
|
104 |
}
|
requirements.txt
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
gradio==4.
|
2 |
azure-storage-file-share==12.11.1
|
3 |
azure-storage-blob
|
4 |
python-dotenv==1.0.0
|
|
|
1 |
+
gradio==4.19.1
|
2 |
azure-storage-file-share==12.11.1
|
3 |
azure-storage-blob
|
4 |
python-dotenv==1.0.0
|
style.css
CHANGED
@@ -319,4 +319,43 @@ span.chatbot > p > img{
|
|
319 |
|
320 |
.card-image > .card-content{
|
321 |
background-color:#f1f7fa !important;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
322 |
}
|
|
|
319 |
|
320 |
.card-image > .card-content{
|
321 |
background-color:#f1f7fa !important;
|
322 |
+
}
|
323 |
+
|
324 |
+
|
325 |
+
|
326 |
+
.tab-nav > button.selected{
|
327 |
+
color:#4b8ec3;
|
328 |
+
font-weight:bold;
|
329 |
+
border:none;
|
330 |
+
}
|
331 |
+
|
332 |
+
.tab-nav{
|
333 |
+
border:none !important;
|
334 |
+
}
|
335 |
+
|
336 |
+
#input-textbox > label > textarea{
|
337 |
+
border-radius:40px;
|
338 |
+
padding-left:30px;
|
339 |
+
resize:none;
|
340 |
+
}
|
341 |
+
|
342 |
+
#input-message > div{
|
343 |
+
border:none;
|
344 |
+
}
|
345 |
+
|
346 |
+
#dropdown-samples{
|
347 |
+
/*! border:none !important; */
|
348 |
+
/*! border-width:0px !important; */
|
349 |
+
background:none !important;
|
350 |
+
|
351 |
+
}
|
352 |
+
|
353 |
+
#dropdown-samples > .container > .wrap{
|
354 |
+
background-color:white;
|
355 |
+
}
|
356 |
+
|
357 |
+
|
358 |
+
#tab-examples > div > .form{
|
359 |
+
border:none;
|
360 |
+
background:none !important;
|
361 |
}
|