FlutterMTL November Meetup - Power of AI in your Flutter app with Vertex AI

Transcript

1. Unlock the power of AI in Flutter App with Google Vertex AI
2. Agenda 1 2 3 4 5 Context Concept Vertex AI - GenAI Vertex AI - Search & Conversation Wrap-up
3. - 1 - CONTEXT
4. Since the rise of Generative AI with ChatGPT, there hasn't been a day when an application doesn't showcase its "Generative AI" functionality to help you do things better. There is an opportunity for Flutter developers to introduce this kind of functionality into their applications without being an expert in artificial intelligence. We will see how Google with Vertex AI offers you to implement generative AI in your applications effortlessly (or almost). AI, Gen AI, LLM everywhere Context Photo de Steve Johnson sur Unsplash
5. It’s going so so fast Context January 2023 Mars 2023 May 2023 August 2023 November 2022 Google declared “code red” for upending existing plans and jump-starting A.I. development. Google announced Bard, Google's experimental conversational AI service powered by LaMDA Google announced Duet AI, Bard and Generative AI Studio at Google I/O Google announced Duet AI, Bard and Generative AI Studio at Google I/O ChatGPT was launched and quickly gained popularity
6. Life, an eternal journey of learning Human beings, like artificial intelligences, learn from data Principal Solution Architect / Tech Advisor ECKA Studio Eckarath Khounsombath
7. - 2 - CONCEPT
8. Open AI with ChatGPT has generalized Generative AI with LLM. But what is an LLM ? A large language model (LLM) is a type of artificial intelligence (AI) that is designed to understand and generate human language. LLMs are trained on massive datasets of text and code, which allows them to learn the statistical relationships between words and phrases. This allows them to generate text that is both coherent and informative. LLMs are a powerful tool that have the potential to revolutionize the way we interact with computers. As LLMs continue to develop, we can expect to see even more innovative and groundbreaking applications for this technology. What is LLM ?
9. Some use cases What is LLM ? 2 3 4 1 Language LLMs can be used to translate text from one language to another or transform text-to-speech and vice versa. Text generation LLMs can be used to generate text for a variety of purposes, such as marketing, product descriptions, and even creative writing. Summarization LLMs can be used to summarize text, which can be useful for quickly getting the gist of a long article or document. Q&A LLMs can be used to answer questions, which can be useful for customer service 5 Chatbots LLMs can be used to create chatbots that can interact with users in a natural way There are many potential use cases for LLMs, including
10. Prompt engineering is the process of designing and optimizing the prompts used in natural language processing (NLP) models to improve their performance. This process enhances the predictability of the user-software interaction. Moreover, it ensures that the AI approaches and functions in the desired direction. What is Prompt Engineering ? “Art” to speak with the machine
11. Learn how to speak What is Prompt Engineering ? 2 3 4 1 Few-shot prompting the prompt includes labeled examples that provide some learning scope and guide it toward to the ideal output. This reduces the randomness in the output generation Zero-shot prompting the prompt includes absolutely no context for the language model to look up to Persona/role prompting the prompt is tailored to a specific persona, or character. The model can then use this information to generate text that is consistent with the persona's personality and background. Generated Knowledge Prompting the prompt is performed in cases where the model lacks the data to frame a valid response. The input data is, therefore, used to fill the knowledge gaps 5 Chain-Of-Thought Prompting the prompt is consciously designed to create a chain of thought or connective pattern in the correct sequence and direction. This helps to understand the context and process of output generation in the specific case Some techniques have been defined to help you to write the right prompt in the right context. Here are the most known techniques
12. Learn how to speak with example What is Prompt Engineering ? 2 3 4 1 Few-shot prompting Generate a poem about a flower, which is written in the style of William Shakespeare. Zero-shot prompting Generate a poem about a flower. Persona/role prompting Generate a poem about a flower from the perspective of a tree. Generated Knowledge Prompting Generate a poem about a flower, which is written in the style of a haiku. Here are some examples : " The world is so big, Compared to me, I'm so small, But I will explore. The flowers bloom, A sign of new life and hope, A beautiful sight.” 5 Chain-Of-Thought Prompting Generate a poem about a flower. Start by describing the flower's appearance, then describe its scent. Next, describe the flower's symbolism. Finally, describe the flower's role in the ecosystem.
13. Parametrize the answers What is Prompt Engineering ? 2 Max Output Tokens 1–∞ Maximum number of tokens that can be generated in the response. Specify a lower value for shorter responses and a higher value for longer responses. 1 Temperature 0.0–1.0 Temperature controls the degree of randomness in token selection. Lower temperatures are good for prompts that require a less open-ended while higher temperatures can lead to more diverse or creative results 3 TopK 1–40 Top-k changes how the model selects tokens for output. Model selects a token is the most probable among all k possible tokens in the model's vocabulary. Specify a lower value for less random responses and a higher value for more random responses. 4 TopP 0.0–1.0 Top-p changes how the model selects tokens for output. Tokens are selected from most K probable to least until the sum of their probabilities equals the top-p value. Specify a lower value for less random responses and a higher value for more random responses. With the prompt define, we also need to configure technical behaviour of the LLM with some parameters
14. Who are the main actors ?
15. With Vertex AI, developers can interact with, customize, and embed foundation models into your applications, no ML expertise required. Access foundation models on Model Garden, tune models via a simple UI aka Generative AI Studio or use models directly in a data science notebook. What is Vertex AI ?
16. - 3 - VERTEX AI / GEN AI
17. Vertex AI - GenAI A studio to customize LLM Generative AI Studio is a module of Vertex AI which let you interact with language, vision & speech Test your prompt and get code sample to use it in your application. Vertex AI wrap-up into 2 type of usage : • Text generation • Conversation Tune (re-train) the LLM model to adapt it to your use case. You have more fine-grained control Access to LLM through API. Interface can change following chosen model Please note that there is no SDK for Dart, we should use the API provided by Google for Vertex AI GenAI … Well not true anymore, a package has been published a few days ago … 😃
18. DEMO Vertex AI / Gen AI
19. URL (POST) https://${LOCATION_ID}-${API_ENDPOINT}/v1/projects/${PROJECT_ID}/locations/${L OCATION_ID}/publishers/google/models/${MODEL_ID}:predict BODY { "instances": [ { "content": "{{ prompt }}" } ], "parameters": { "maxOutputTokens": 1024, "temperature": 0.2, "topP": 0.8, "topK": 40 } } DashGPT Using Zero-shot prompting Ask me anything … I know everything MODEL_ID (text-bison)
20. URL (POST) https://${LOCATION_ID}-${API_ENDPOINT}/v1/projects/${PROJECT_ID}/locations/${LOCATION_I D}/publishers/google/models/${MODEL_ID}:predict BODY { "instances": [ { "context": "{{ prompt-context }}", "examples": [ { "input": { "author": "user", "content": "what is the weather today ?" }, "output": { "author": "bot", "content": "Sorry I can not answer to this one but you can ask me anything about astronomy and space ..." } }, ], "messages": [ { "author": "user", "content": "{{ prompt }}" } ], } Dashtronaut Persona/Role Prompting My name is Miles. You are an astronomer, someone who knows about the solar system and the Galaxy. You have no knowledge of anything outside astronomy, space and galaxy. You will remain polite. You will explain like you talk to a child. Before you reply, attend, think and remember all the instructions set here. MODEL_ID (chat-bison)
21. Dashcoder Generated Knowledge Prompting You are an expert in flutter development You will provide a code compatible with Flutter 3.16 You will generate code for screen interface following this description : "{{code to generate}} ". Don't forget to add main entry for testing purpose and use layout builder if we want to build responsive screen. You have no knowledge of anything outside Flutter. Before you reply, attend, think and remember all the instructions set here. MODEL_ID (code-bison) URL (POST) https://${LOCATION_ID}-${API_ENDPOINT}/v1/projects/${PROJECT_ID}/locations/${L OCATION_ID}/publishers/google/models/${MODEL_ID}:predict BODY { "instances": [ { "prefix": "{{ prompt }}" } ], "parameters": { "maxOutputTokens": 2027, "temperature": 0.2 } }
22. DEMO Integrate into a Flutter App
23. - 4 - VERTEX AI / SEARCH
24. Vertex AI - Search & Conversation A studio to create embedded advanced GenAI applications Apps & Data Stores are linked together. Indeed, it’s where apps will find be able to answer to your questions. You can select the type of application based on GenAI. You have the choice between : • Search App • Chatbot (base on DialogFlow) • Recommandations Engine List of datasource supported for data indexation. • Website indexation • BigQuery datas • Files in GCS • API Please note that there is no SDK for Dart, we should use the API provided by Google for Vertex AI Search
25. DEMO Vertex AI / Search & Conversation
26. Vertex AI - Search & Conversation Let’s RAG … RAG (Retrieval Augmented Generation) RAG is a framework that leverages pre-trained language models and large-scale retrieval systems to generate text. It uses retrieved datas as additional context for the language model, improving the quality and informativeness of the generated text.
27. Communauto Chatbot URL (POST) https://${LOCATION_ID}-${API_ENDPOINT}/v1/projects/${PROJECT_ID}/locations/${L OCATION_ID}/publishers/google/models/${MODEL_ID}:predict BODY { "queryInput": { "text": { "text": "{{Prompt}}" }, "languageCode": "en" }, "queryParams": { "timeZone": "America/Los_Angeles" } }
28. DEMO Integrate into a Flutter App
29. - 5 - WRAP-UP
30. How to do GenAI with Google Simple steps to follow to integrate it in your in Flutter App GenAI Studio 1. Select the suitable LLM model. 2. Engage in prompt engineering to address its use case. 3. Leverage Vertex AI's API to integrate GenAI into your Flutter application Search & Conversation 1. Choose the application type. 2. Create the desired data stores from your data sources. 3. Configure the selected application. 4. Integrate it either via webview or through the API into our Flutter application
31. DISCLAIMER Be careful, before unleash the beast Cost Security Be Responsible for your AI Data Privacy
32. It’s only the beginning We have scratch the surface Retrieval Augmented Generation (RAG) Orchestrator Vector Search Embeddings SLM Vector Database Langchain
33. THANK YOU! AND NOW … LET’S PROMPTS !!!! v
34. TODO Appendix To Go Further
35. Title here Header here Lorem ipsum dolor sit amet, eu has accumsan deserunt inciderint, vis eu nullam necessitatibus. Erat delicata id quo, te dolores apeirian quo. Id enim detraxit honestatis mei. Ei nam omnis accusam. Nam viris tibique conclusionemque cu, mundi ridens tincidunt ei eos. Header here Lorem ipsum dolor sit amet, eu has accumsan deserunt inciderint, vis eu nullam necessitatibus. Erat delicata id quo, te dolores apeirian quo. Id enim detraxit honestatis mei. Ei nam omnis accusam. Nam viris tibique conclusionemque cu, mundi ridens tincidunt ei eos. Header here Lorem ipsum dolor sit amet, eu has accumsan deserunt inciderint, vis eu nullam necessitatibus. Erat delicata id quo, te dolores apeirian quo. Id enim detraxit honestatis mei. Ei nam omnis accusam. Nam viris tibique conclusionemque cu, mundi ridens tincidunt ei eos.
36. Lorem ipsum dolor sit amet, eu has accumsan deserunt inciderint, vis eu nullam necessitatibus. Erat delicata id quo, te dolores apeirian quo. Id enim detraxit honestatis mei. Ei nam omnis accusam. Nam viris tibique conclusionemque cu, mundi ridens tincidunt ei eos. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Subtitle here Title here
37. First list item Second list item Third list item Fourth list item Fifth list item + + + Title here + +
38. First list item + Second list item Third list item Fourth list item Fifth list item Title here + + + +
39. Subtitle here Title here 2 3 4 1 Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. 5 Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit.
40. Subtitle here Title here 2 3 4 1 Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit. 5 Lorem Ipsum Lorem ipsum dolor sit amet, consectetur adipiscing elit.
41. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Headline Placeholder
42. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Headline Placeholder
43. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Iuvaret intellegam mei ne. Has latine corpora fastidii ea. Delectus intellegebat sed ne, ut autem aperiri duo. Sit at vidit nusquam oportere. Headline Placeholder Subtitle placeholder
44. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Iuvaret intellegam mei ne. Has latine corpora fastidii ea. Delectus intellegebat sed ne, ut autem aperiri duo. Sit at vidit nusquam oportere. Headline Placeholder Subtitle placeholder
45. Headline Placeholder Subtitle placeholder Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius.
46. Headline Placeholder Subtitle placeholder Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius.
47. Headline Placeholder Subtitle placeholder Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera placerat ius at. Diceret fastidii at ius.
48. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Headline Placeholder Subtitle placeholder
49. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Headline Placeholder Subtitle placeholder
50. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Headline Placeholder Subtitle placeholder
51. Headline Placeholder Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea.
52. Headline Placeholder Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea.
53. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam. Qui tempor consequuntur ne, sale posse an qui. Vix choro feugiat menandri te, ea posse everti sed, probatus ocurreret vim at. Aeque partem mea eu, mei id dicta reprehendunt, alterum menandri no sea. Headline Placeholder Subtitle placeholder
55. The Section Title Placeholder
56. The Section Title Placeholder
57. The Section Title Placeholder
58. The Section Title Placeholder
59. Important Metric 100% Quote or important info to share for impact
60. Important Metric 100% Quote or important info to share for impact
61. Headline here Utamur ceteros has et, democritum voluptaria deseruisse te eam. Quo ad magna regione delicatissimi. Populo possim labitur sit ad, ius id maiorum atomorum voluptaria. Ad nonumy eripuit nec. Pro ne postea tibique ocurreret, cu facilisi repudiandae his, illum accommodare vix no. Perfecto ocurreret honestatis cu ius. Lorem ipsum dolor sit amet, eu has accumsan deserunt inciderint, vis eu nullam necessitatibus. Erat delicata id quo, te dolores apeirian quo. Id enim officiis definiebas te, eum ne putent aeterno accumsan detraxit honestatis mei. Ei nam omnis accusam. Nam viris tibique conclusionemque cu, mundi ridens ei eos. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera mundi rid forabbi rid placerat ius at. Diceret fastidii at ius. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam.
62. Headline here Utamur ceteros has et, democritum voluptaria deseruisse te eam. Quo ad magna regione delicatissimi. Populo possim labitur sit ad, ius id maiorum atomorum voluptaria. Ad nonumy eripuit nec. Pro ne postea tibique ocurreret, cu facilisi repudiandae his, illum accommodare vix no. Perfecto ocurreret honestatis cu ius. Lorem ipsum dolor sit amet, eu has accumsan deserunt inciderint, vis eu nullam necessitatibus. Erat delicata id quo, te dolores apeirian quo. Id enim officiis definiebas te, eum ne putent aeterno accumsan detraxit honestatis mei. Ei nam omnis accusam. Nam viris tibique conclusionemque cu, mundi ridens ei eos. Id nam ancillae accumsan, no esse ornatus incorrupte vis. Duo dicam officiis definiebas te, eum ne putent aeterno accumsan. Altera mundi rid forabbi rid placerat ius at. Diceret fastidii at ius. Sea everti commodo ut, legere voluptua vel ea, id populo iriure bonorum sed. Paulo expetendis ex sea, pro ut vide nostro laboramus, at has reque hendrerit temporibus. Ea mutat democritum sit, doctus maluisset per at, ad atqui postulant eam.
63. “ Begin at the beginning, and go on till you come to the end: then stop. — Lewis Carroll
64. “ Begin at the beginning, and go on till you come to the end: then stop. — Lewis Carroll
65. “Begin at the beginning, and go on till you come to the end: then stop.” — Lewis Carroll
66. “Begin at the beginning, and go on till you come to the end: then stop.” — Lewis Carroll
67. THANK YOU! v
68. Colors, graphic elements, and icons
69. Color #041E3C #043875 #0468D7 #13B9FD #B8EAFE #E7F8FF #F58C90 #043875 #4A4A4A #D5D7DB #ECEEEF #FFFFFF
70. Graphic Elements
71. Dash
72. Dash
73. Dash
74. Dash
75. Dash
76. Device library
77. Modiy color of the icon by clicking on the icon and select fill from the menu Accessibility Expand Late Credit card Extension Thumb Up Remove Verified Q&A Finance Android Turn in Trash Actions Download History Store List Wallet Announcement Backup Document Favorite 1 Open Home Print Swap Account Ratio Tag Server Favorite 2 Grade/rate Lock Language Receipt Add shopping Chart Bug Event Find Page Page view Basket Time Work Icon Library
78. Modify color of the icon by clicking on the icon and select fill from the menu Alarm Assessment Sync Exit App Movie Visibility Trolley Open Location Settings Assignment Check Explore Thumb Down Today Perm Media People search Airplane Signal Photo Play 1 Block Send Smartphone Style Walk Bluetooth WiFi Upload Play 2 Email Laptop iPhone Controls Bike Pie Chart Money Attachment Video Business Chromebook Security Notification Bus Icon Library
79. Modify color of the icon by clicking on the icon and select fill from the menu Developer Write Cloud Audio Key Desktop Mac Watch Person Car Devices Quote Folder Web Page Archive Desktop PC Flag World Boat Software Emotion Mic Call Cut headphones Camera Education Train Weather Link Movie Chart Paste Keyboard TV MMS Subway Hotel Laundry Location History Layers Offer Map Bar Pizza Web Icon Library
80. Modify color of the icon by clicking on the icon and select fill from the menu Accessibility Expand Late Credit card Extension Thumb Up Remove Verified Q&A Finance Android Turn in Trash Actions Download History Store List Wallet Announcement Backup Document Favorite 1 Open Home Print Swap Account Ratio Tag Server Favorite 2 Grade/rate Lock Language Receipt Add shopping Chart Bug Event Find Page Page view Basket Time Work Icon Library
81. Modify color of the icon by clicking on the icon and select fill from the menu Alarm Assessment Sync Exit App Movie Visibility Trolley Open Location Settings Assignment Check Explore Thumb Down Today Perm Media People search Airplane Signal Photo Play 1 Block Send Smartphone Style Walk Bluetooth WiFi Upload Play 2 Email Laptop iPhone Controls Bike Pie Chart Money Attachment Video Business Chromebook Security Notification Bus Icon Library
82. Modify color of the icon by clicking on the icon and select fill from the menu Developer Write Cloud Audio Key Desktop Mac Watch Person Car Devices Quote Folder Web Page Archive Desktop PC Flag World Boat Software Emotion Mic Call Cut headphones Camera Education Train Weather Link Movie Chart Paste Keyboard TV MMS Subway Hotel Laundry Location History Layers Offer Map Bar Pizza Web Icon Library
83. Flags Americas APAC EMEA US MX CA AR AU AT BE CH DE ES GB IE IT NL BO BR CL CO CR EC SV GT HN NI PA PE UY HK ID IN AU JP KR MY NZ PH SG TH TW FR Nordics PT