Releases: instill-ai/instill-core
v0.16.0-alpha
Support for Pinecone, BigQuery, and Google Cloud Storage Connectors
We’re taking VDP to the next level with new Connectors. Other than that, we’ve made some small but important improvements to the console and UI/UX.
Highlights:
- Pinecone Connector: Long-term memory for AI.
- BigQuery Connector: Access a large volume of data.
- Google Cloud Storage Connector: Access a large volume of data.
- Console UX improvements: Making our users’ lives easier.
Incorporate VDP into your workflow effortlessly, whether on the cloud or through self-hosting. Explore the options below:
Pinecone Connector
Now you have the power of embeddings with Pinecone. Pinecone provides optimal storage and querying capabilities; ensuring performance, scalability, and flexibility for embeddings. You can now enhance your AI apps with advanced features such as semantic information retrieval and long-term memory.
BigQuery and Google Cloud Storage Connectors
The full potential of AI can be realized when it is integrated with your existing databases and the vast amount of information available online.
We simplify the process of connecting to various data sources such as warehouses, databases, APIs, and vector databases.
Now, we’re introducing these connectors for you to connect to the BigQuery data warehouse and Google Cloud Storage so that you can connect to your own data.
Numerous Console UX improvements
Now you have better control of your pipelines, not only in the form of version control but also pipeline duplication (forking).
And, many more improvements, we focused on the finer details this time!
What's next?
We’ve also worked hard on the backend to be able to provide some even more exciting features in the next release!
We highly value your feedback, so feel free to initiate a conversation on GitHub Discussions or join us in the #vdp channel on Discord.
Your insights drive us forward ❤️!
v0.15.0-alpha
Introducing pipeline templates
VDP is now at a point of a solid (alpha) foundation. This is a major release packed with features and updates which will allow you to not only build unstructured data pipelines and AI apps but also build them with more flexibility, control, and ease.
Highlights:
- New onboarding flow + Templates: Get started faster.
- Version control: Build better.
- Pipeline and App Sharing + Cloning: For teams and community sharing.
- (Simple) authentication for Instill Core: Increased security.
Incorporate VDP into your workflow effortlessly, whether on the cloud or through self-hosting. Explore the options below:
Templates, Templates, Templates
Introducing VDP templates
This is our favourite new feature! We’re super excited to share these valuable use cases with you. Now, you don’t have to start from scratch, use any of these templates to solve real-world problems.
Login to check out the first few ready-to-go templates here.
Version Control
Pipeline version control
With these features, you can gain more flexibility and control. Version Control allows users to track changes in their pipelines and apps over time. Not only that, but users can use different versions of the same pipeline; for backup and recovery, A/B testing and etc.
Pipeline Sharing + Cloning
Sharing your pipelines
You can share your creations using a link or choosing to make the creation public. Another user can then view the shared pipeline or app and clone it to their workspace. We believe this will reduce barriers and enable the usage of complex unstructured data pipelines and AI apps by a larger audience.
(Simple) Authentication for Instill Core
We’ve introduced a single user login feature in Instill Core. With this feature, there is now basic authentication to protect your data resource.
We’ve also integrated API tokens into Instill Core to enhance your authentication and authorization processes, granting users and applications secure access to your pipelines.
What's next?
Stay tuned for these exciting updates as we work on improving the user-friendliness and value of VDP. We highly value your feedback, so feel free to initiate a conversation on GitHub Discussions or join us in the #vdp channel on Discord.
Your insights drive us forward ❤️!
Some technical stuff:
Bug Fixes
Miscellaneous Chores
- release: release v0.15.0-alpha (19fe7f8)
v0.14.0-alpha
🎉 Highlights
We've totally reimagined the user experience when it comes to building a pipeline, check it out!
Plus, we now support the following:
- Pinecone as a data connector
- Hugging Face as AI Connector
- Testing of pipeline within the canvas
- Array UI in test mode
☁️ Try Instill Cloud for FREE
Instill Cloud is now available as a fully managed public cloud service! As the easiest and most reliable way to build the backbone of your AI-first applications, it comes with all features available for FREE during the Open Alpha phase. We eagerly await your valuable feedback as we strive to make Instill Cloud the ultimate AI and Data tooling platform 🙌!
🎯 What's next?
We'll continue to work on exiting upgrades to help you customize your pipelines even further, making it super easy to work with your app data in any form you prefer. Stay tuned!
We'll also introduce functionality that allows new users to get started with building a pipeline instantly, along with user experience improvements.
Stay connected for these remarkable updates as we continue to enhance VDP's user-friendliness and potency. Your feedback is invaluable – don't hesitate to start a discussion on GitHub Discussions or join us in the #vdp channel on Discord.
Your insights drive us forward ❤️!
v0.13.0-alpha
🎉 Highlights
The OpenAI AI Connector has expanded its capabilities and now supports the "Audio Recognition" task in addition to the existing "Text Generation" and "Text Embeddings" tasks.
☁️ Try Instill Cloud for FREE
Instill Cloud is now available as a fully managed public cloud service! As the easiest and most reliable way to build the backbone of your AI-first applications, it comes with all features available for FREE during the Open Alpha phase. We eagerly await your valuable feedback as we strive to make Instill Cloud the ultimate AI and Data tooling platform 🙌!
🔮 OpenAI AI Connector now supports "Audio Recognition" AI task
To utilize this capability, simply create an OpenAI AI Connector and choose the "Audio Recognition" task type. Alternatively, you can use the pre-created connector instill-openai-speech-recognition-whisper-1 available on Instill Cloud!
The OpenAI AI Connector supports the "Audio Recognition" AI task
Create a simple pipeline with OpenAI AI Connector to transcribe your audio to text
To trigger the audio recognition, create a simple pipeline and submit a POST request with the audio file URL:
curl -X POST 'https://api-vdp.instill.tech/v1alpha/pipelines/transcriber/trigger' \
--header 'Authorization: Bearer <INSTILL_CLOUD_API_TOKEN>' \
--data-raw '{
"inputs": [
{
"audios": [
{
"url": "https://artifacts.instill.tech/audios/vdp-release-v0.13.0-alpha.m4a"
}
]
}
]
}'Or, if you prefer, encode your audio file in base64 format:
curl -X POST 'https://api-vdp.instill.tech/v1alpha/pipelines/transcriber/trigger' \
--header 'Authorization: Bearer <INSTILL_CLOUD_API_TOKEN>' \
--data-raw '{
"inputs": [
{
"audios": [
{
"blob": "<base64-encoded-audio-file>"
}
]
}
]
}'The transcript will be available in the "texts" field of the response. Excited to know more? Check out the details in the "What's next" section below 👇
All updates are applied to Instill Cloud.
🎯 What's next?
Here's the exciting transcript from the audio file mentioned above:
Hello everyone, calling all AI application builders, we are about to bring a major upgrade to VDP's no-code pipeline builder. It's gonna be fantastic. This upgrade will let you customize your VDP pipelines just the way you like, making it super easy to work with your app data in any form you prefer. Stay tuned!
That's right! The upcoming VDP version is packed with incredible features, designed to enhance your user experience:
-
Empower users: We aim to provide you with a seamless pipeline-building experience, reducing the reliance on documentation. With intuitive components, you'll be able to create pipelines effortlessly
-
Streamlined UI: Recognizing the importance of clear data field mapping and visualized data flows, we've revamped the UI to offer a simplified and lucid representation. This ensures an enhanced and user-friendly experience
Stay connected for these remarkable updates as we continue to enhance VDP's user-friendliness and potency. Your feedback is invaluable – don't hesitate to start a discussion on GitHub Discussions or join us in the #vdp channel on Discord.
Your insights drive us forward ❤️!
v0.12.0-alpha
🎉 Highlights
We're excited to expand our Connector offerings! With this latest update, we have added support for additional Connectors, making it even easier to integrate and customize your AI-first applications.
☁️ Try Instill Cloud for FREE
Instill Cloud is now available as a fully managed public cloud service! As the easiest and most reliable way to build the backbone of your AI-first applications, it comes with all features available for FREE during the Open Alpha phase. We eagerly await your valuable feedback as we strive to make Instill Cloud the ultimate AI and Data tooling platform 🙌!
✨ Manipulate components in the pipeline builder in Console
We are thrilled to introduce a powerful new feature in the Console's pipeline builder. Now, you can effortlessly manipulate components within the pipeline, granting you full control over how data flows through the process.
Here's an example: Imagine building two pipelines that generate images with text prompts using a Stability AI Connector. The key distinction between these pipelines lies in the metadata field of the trigger component, which is connected to the response metadata field. When triggering both pipelines with the same request:
curl -X POST 'https://api-vdp.instill.tech/v1alpha/pipelines/<pipeline-id>/trigger' \
--header 'Authorization: Bearer <INSTILL_CLOUD_API_TOKEN>' \
--data-raw '{
"inputs": [
{
"texts": [
"Ragdoll cat king wearing a golden crown, intricate, elegant, highly detailed, centered, digital painting, artstation, concept art, smooth, sharp focus, illustration, artgerm, Tomasz Alen Kopera, Peter Mohrbacher, donato giancola, Joseph Christian Leyendecker, WLOP, Boris Vallejo"
],
"metadata": {
"samples": 1,
"seed": 143938098,
"sampler": "K_EULER_ANCESTRAL",
"cfg_scale": 7.5,
"steps": 30,
}
}
]
}'
Pipeline 1 (No metadata connections)
{
"outputs": [
{
"data_mapping_index": "01H5TS75D4CP39WTG28Z6JG2XD",
"texts": [],
"images": [
{
"blob": "< generated image >"
}
],
"structured_data": null,
"metadata": null
}
]
}
Pipeline 2 (With metadata connections)
{
"outputs": [
{
"data_mapping_index": "01H5TRWBH60DTYTR0W5JA8DCAB",
"texts": [],
"images": [
{
"blob": "< generated image >"
}
],
"structured_data": null,
"metadata": {
"cfg_scale": 7.5,
"pipeline": {
"id": "<pipeline-id>",
"owner": "users/<uuid>"
"trigger_id": "<uuid>",
"uid": "<uuid>"
},
"sampler": "K_EULER_ANCESTRAL",
"samples": 1,
"seed": 143938098,
"steps": 30
}
}
]
}The response of pipeline 2 demonstrates that the input metadata seamlessly passes all the way to the response.
This feature empowers you to achieve custom data flows tailored to your specific requirements.
✨ New Metric Dashboard for observability
Our latest update brings you a brand-new Metric Dashboard in the Console, providing enhanced observability into pipeline trigger usage. You can now easily view pipeline trigger metrics for each pipeline, filter triggers by date range and pipeline state, and see the number of successful and errored triggers.
To delve deeper into trigger logs for a specific pipeline, simply click on the corresponding pipeline ID in the table. You will be directed to a detailed pipeline trigger table, displaying trigger time, status, and duration for each trigger.
We have more exciting metrics planned, including Connector usage, giving you a comprehensive overview of the usage of all your favorite Connectors.
🔮 Introducing new OpenAI AI Connector
OpenAI's LLMs redefine language processing with unparalleled power. Harnessing the might of GPT-3.5 and the latest GPT-4, these models exhibit extraordinary capabilities in understanding, generating, and transforming text. From creative storytelling to practical problem-solving, OpenAI LLMs empower businesses and individuals to conquer new frontiers in the realm of natural language understanding. For all the popular use cases, we can do it with VDP, and even more, thanks to the OpenAI integration!
Create an OpenAI AI Connector to connect to models like GPT-4 and GPT-3.5 for "Text Generation" and "Text Embeddings" tasks.
OpenAI <> VDP
🔮 Stability AI AI Connector now supports "Image to Image" AI task
Exciting news! The Stability AI AI Connector now supports the "Image to Image" AI task, in addition to the existing "Text to Image" AI task. This enhancement allows you to generate new images from existing ones with ease.
To utilize this feature, when creating a Stability AI AI Connector, select the "Image to Image" task type and create a simple pipeline.
Create a Stability AI AI Connector for "Image to Image" task
Create a simple pipeline with Stability AI AI Connector
Trigger it by filling both the "texts," "images," and "metadata" fields.
curl -X POST 'https://api-vdp.instill.tech/v1alpha/pipelines/aigc/trigger' \
--header 'Authorization: Bearer <INSTILL_CLOUD_API_TOKEN>' \
--data-raw '{
"inputs": [
{
"texts": [
"animated drawing of dogs sit in snow"
],
"images": [
{
"url": "https://artifacts.instill.tech/imgs/dog.jpg"
}
],
"metadata": {
"weights": [0.4],
"init_image_mode": "IMAGE_STRENGTH",
"image_strength": 0.6,
"samples": 1,
"seed": 6987,
"cfg_scale": 7.5,
"steps": 30
}
}
]
}'The response will include the generated image in Base64 format.
The Stability AI AI Connector supports the "Image to Image" AI task
All updates are applied to Instill Cloud.
🎯 What's next?
We are thrilled to share our exciting plans for VDP in the upcoming releases. Our primary focus will be on revamping the no-code pipeline builder with a component-oriented design.
The goals of this redesign are to:
-
Empower users: Our aim is to provide users with a seamless pipeline-building experience, reducing the need to rely heavily on documentation. With intuitive components, you'll be able to create pipelines effortlessly.
-
Simplify UI: We understand the importance of clarity when mapping data fields and visualizing data flows. The redesigned UI will offer a simple and clear representation of these aspects, ensuring an enhanced user experience.
Stay tuned for these amazing updates as we strive to make VDP even more user-friendly and powerful. As always, we value your feedback, so feel free to create a topic in Discussions, or meet us in the #vdp channel on Discord.
Happy building ❤️!
v0.11.0-alpha
🎉 Highlights
We're excited to announce the latest updates including new AI Connectors and a No-Code Drag-and-Drop Pipeline Builder in Console!
☁️ Try Instill Cloud for FREE
Instill Cloud is a fully-managed public cloud service. It is the easiest and most reliable way to build the backbone of your AI-first applications. The service is currently in Open Alpha - with all features for FREE. We eagerly await your feedback as we strive to make Instill Cloud the best AI and Data tooling can be 🙌!
✨ Say hello to our no-code Drag-and-Drop Pipeline Builder in Console
Get ready to experience sheer convenience with our newly revamped Console! We've gone the extra mile to bring you a slick no-code Drag-and-Drop Pipeline Builder.
With this exciting feature, building your data pipelines has never been easier. Simply drag and drop the connector components onto the canvas, connect them together to form a pipeline, and watch the magic happen! Visualise how your data flows through the pipeline and undergoes processing. You don't need to be a tech-savvy expert to navigate this tool - it's designed to empower everyone to become a data pro.
We're working tirelessly to continuously improve the Console, ensuring it provides the best user experience for building and monitoring your unstructured data pipelines. Our goal is to make the process seamless and intuitive, so you can focus on getting the most out of your data.
✨ Introduce new AI and Blockchain connectors into VDP
In addition to decoupling model deployment, we have introduced two new connector types in this release: AI Connector and Blockchain Connector.
Introducing new VDP connectors
The AI Connector allows seamless integration of AI models from various sources, including both model providers (e.g., 1st-party ModelOps platform Instill Model (read more below) and 3rd-party vendors like Stability AI, OpenAI and Hugging Face) and in-house model serving platforms. It replaces the deprecated Model component and processes unstructured data into analysable or meaningful data representations. AI connectors are maintained in the connector-ai repository.
📔 Explore the documentation
Furthermore, we have introduced the Blockchain Connector, which facilitates integration with 3rd-party Blockchain services. It enables the conversion of data into Web3 assets. For instance, you can upload AI-generated or AI-assisted content on the blockchain to ensure transparency, trust, and establish a tamper-proof record of the content's origin, creation, and subsequent modifications. Blockchain connectors are maintained in the connector-blockchain repository.
📔 Discover the documentation
Tutorial for AIGC x Web3 Pipeline
Ready to dive into the cutting-edge world of generative AI and blockchain? Check out our latest tutorial: From Generative AI to Blockchain: Building an AIGC x Web3 Pipeline with Instill VDP. This comprehensive guide will walk you through building a VDP pipeline that utilizes Stability AI's SDXL generation model and Numbers Protocol's blockchain technology to register your AI-generated data as Web3 assets.
Embrace the future of AI and blockchain integration with this exciting tutorial!
✨ Re-structured Instill products
We understand that model deployment has been a challenging aspect while using VDP. Previously, users had to download a large Triton docker image and ensure model compatibility with Triton to utilize VDP. To address this, we have reimagined VDP and introduced a standalone service called Instill Model. With this restructuring, we decoupled model deployment from VDP and made it lightweight.
As we mentioned above, we have introduced the AI connector into VDP, enabling connections to various AI models from different sources such as our 1st-party model platform, Instill Model, and 3rd-party providers like Stability AI, OpenAI, and Hugging Face.
As a result, we now have Instill Core, consisting of three products:
- Base: this includes essential services like user management, databases, and observability tools required to launch VDP and Model.
- VDP: a no-/low-code tool that focuses on building versatile data pipelines for unstructured data.
- Model: your go-to for ModelOps to import and serve ML model deployment, ready to dish out some serious predictions at scale.
To ensure efficiency at scaling, our products adopt a microservice architecture design. We aim to make each product standalone, enabling independent debugging, deployment, monitoring, and scaling. For more information, please review the updated system infrastructure.
Instill Core System Architecture
✨ Standardise the connector data format
Our primary goal with VDP is to streamline the end-to-end unstructured data flow. To achieve this, we have standardized the connector data input and output format. This standardization allows connectors in a pipeline to process and pass data efficiently. It also ensures that VDP produces a stream of data with a consistent format, which can be easily utilized in data integration or data pipelines.
Here is the standardised connector data input/output format:
// Standardised connector data input/output format
// - "texts": an array containing your text data
// - "images": an array containing your image data
// - "structured_data": a JSON object containing well-structured data
// - "metadata": a JSON object containing metadata for the above data
{
"texts": []
"images": [],
"metadata": {}
"structured_data": {}
"metadata": {}
}
🅱️ API breaking changes
We would like to highlight the following breaking changes to our API:
- Deprecation of model-related API endpoints: in order to support the decoupling of the model deployment feature from VDP, we have deprecated the model-related API endpoints (
GET|POST|PATCH|DELETE /v1alpha/models/... - Consolidation of connector endpoints: with the introduction of more connector types (AI, Blockchain and future VectorDB connectors), we have simplified the API design by consolidating all connectors under a single set of endpoints (
/v1alpha/connectors/...). - Update to the pipeline data format: We have updated the pipeline data payload format to trigger a pipeline. This change aligns with the standardised connector data format.
// Batched data inputs to trigger a pipeline
// - "texts": an array containing your text data in string format to be processed
// - "images": an array containing your image data in URL or base64 format to be processed
// - "structured_data": a JSON object containing well-structured data to be processed
// - "metadata": a JSON object containing metadata for the above data
{
"inputs": [
{
"texts": [...],
"images": [
{
"url": <remote-image-url>
},
{
"blob": <base64-image-string>
}
...
],
"metadata": {...}
}
]
}
We apologize for any inconvenience caused by these breaking changes. Please make the necessary adjustments to your implementation accordingly.
All updates are applied to Instill Cloud.
🎯 What's next?
We have exciting plans for VDP in the upcoming releases. Our roadmap includes support for additional AI connectors, such as OpenAI and Hugging Face. We also aim to introduce a new connector type called VectorDB Connector (e.g., Pinecone and Milvus) to cater to new AI tasks involving Embeddings.
To encourage community contributions, we will introduce a VDP connector framework that simplifies the process of creating and adding new connectors. If you have ideas for new connectors, feel free to create a topic in Discussions, or request them in the #vdp channel on Discord.
The introduction of these new connectors opens up significant opportunities for VDP to integrate with various services, enhancing unstructured data processing workflows for your specific use cases.
v0.10.2-alpha
Product Updates
- adapt the new connector definition
All updates are applied to Instill Cloud.
Console (0.24.1-alpha)
Miscellaneous
v0.10.1-alpha
Product Updates
- Support manually testing connector connection on Console
- New Console topbar and sidebar to lay the foundation for UI revamp
- Improve logging format and consistency across VDP services
All updates are applied to Instill Cloud.
Console (0.24.0-alpha)
Features
- add test connector connection button (#422) (409086f)
- style: update console with new sidebar and topbar design (#425) (64cd846)
Bug Fixes
v0.10.0-alpha
🎉 Highlights
New pipeline recipe, telemetry integrations and license updates
Please refer to the following sections for more details.
🅱️ [Breaking change] New pipeline recipe
In this release, we have adopted a new pipeline recipe format (breaking change) to lay the foundation to support more complex and flexible pipelines in the future.
A recipe describes a pipeline resource consisting of components including:
- source component - where the pipeline starts to ingest unstructured data to be processed
- model components - several deployed AI models to process the ingested unstructured data in parallel
- destination components - several destinations to send the processed outputs in parallel
{
"version": "v1alpha",
"components": [
{
"id": <component-id>,
"resource_name": <source-connector-resource-name>
},
{
"id": <component-id>,
"resource_name": <model-resource-name>
},
...
{
"id": <component-id>,
"resource_name": <destination-connector-resource-name>
}
]
}
👉 Check out the documentation here
🅱️ [Breaking change] Trigger endpoint updates
The old triggering endpoint /trigger was retired and replaced with fine-grained endpoints based on the pipeline mode:
- the trigger endpoint for a
SYNCpipeline is/triggerSync - the trigger endpoint for an
ASYNCpipeline/triggerAsync
New telemetry integrations
Observability is critical for distributed microservice architecture. Through OpenTelemetry, we can generate, collect and export metrics, logs and traces to help analyse VDP’s performance and behaviour.
How to enable the VDP observability stack?
You can enable the observability stack by setting ENABLE_OBSERVE=true in the .env file. The following telemetry tools are supported now.
- Jaeger (
localhost:16686): OpenTelemetry allows us to export spans to Jaeger. Use Jaeger when you want to debug the complete flow of a request through the VDP services. - InfluxDB (
localhost: 8086, username:admin, password:password): detailed metrics are sent to InfluxDB for monitoring, and are imported into the Grafana dashboard - Grafana (
localhost:3002, username:admin, password:admin): the Grafana dashboard visualises the metrics to monitor the performance and anomalies of VDP services - Prometheus (
localhost:9090): VDP exports metrics likevdp_pipeline_sync_trigger_counter_total(total number of triggers from SYNC pipelines) andvdp_pipeline_async_trigger_counter_total(total number of triggers from ASYNC pipelines) to Prometheus.
License updates
From this release, VDP will adopt multiple licenses.
- VDP Protocol: open-source and available under the MIT License
- VDP Connectors (including data source connectors, data destination connectors and model connectors): open-source and available under the MIT License
- UI, API and services of VDP: source available under the Elastic License 2.0 (ELv2)
- Instill Cloud (fully-managed public cloud service for VDP): closed-source
Why are VDP connectors open source?
We want to build VDP as an open platform that integrates with other fantastic tooling and services. And the best way to build these integrations is via open source, whether it is a data connector connecting data to a warehouse, application, or database, or a model connector connecting models from self-hosted or 3rd party platforms. This also benefits the projects that want to integrate with these connectors.
Why did VDP services adopt ELv2?
Our mission is to make AI accessible to everyone. The best way to achieve this is to make VDP free to use and source available to everyone, while ensuring we safely create a sustainable business. Therefore, we chose ELv2 since it is very permissive. On the one hand, this allows us to provide users with free access to our source code, the permission to modify it. On the other hand, we've released Instill Cloud, a fully-managed public cloud service built upon VDP that offers additional features. With the license, we don't have to worry about our project being taken by some other companies for monetization.
👉 Check the License FAQ here.
Console 0.23.0-alpha
Features
- [INS-518] update dummy model for testing (#419) (b6a3745)
- [INS-751] adapt new pipeline recipe and update trigger snippet (#417) (61809bb)
- stabilize test related to the pipeline details page (#414) (9f72427)
- update docs link at sidebar (#416) (51103f9)
Bug Fixes
v0.9.3-alpha
🎉 Launching Instill Cloud, a fully managed cloud service that provides you with all the power of VDP, without managing infrastructure
We are thrilled to announce the long-awaited release of Instill Cloud, the fully-managed cloud service for Versatile Data Pipeline (VDP), now available in Open Alpha.
Our goal in developing VDP was to make it accessible to everyone. With a single line, you can launch VDP on local machines without the need to build from scratch. However, we wanted to go even further. We understand that deploying, scaling, and managing infrastructure for unstructured data pipelines can be daunting, especially for those who are not tech-savvy or have limited resources. These challenges are real and not to be taken lightly.
The exciting news is that you now have the option to utilize Instill Cloud and experience the full power of VDP without any hassle. Currently in Open Alpha, Instill Cloud offers the following features:
- Support for SYNC and ASYNC pipelines
- Free pipeline triggers
- Free access to pre-trained ML models
- Unlimited API requests
👉 Visit Instill Cloud to sign up and get started today. We eagerly await your feedback as we strive to make Instill Cloud the best it can be 🙌!
🎯 What's next?
We are actively working on enhancing the user experience by rapidly refactoring VDP and Instill Cloud. Here are some important features on the horizon:
- Supporting flexible and complex pipelines through a DAG design
- Decoupling the model deployment feature from VDP and introducing a standalone ModelOps platform called Instill Model, providing a more stable model inference experience
- Introducing the Model Connector to connect AI models from different sources, whether they are third-party model providers or in-house model serving platforms. This will unlock numerous opportunities for VDP to integrate with popular open model providers like OpenAI, Stability AI, Hugging Face, and, of course, Instill Model
- Expanding support for more data connectors, including popular vector dbs
For more details, please refer to our roadmap.
Model-backend (v0.16.4-alpha)
Bug Fixes
- create single triton client (instill-ai/model-backend#357) (8dedf5d)
Controller (v0.1.4-alpha)
Bug Fixes
- model,connector,pipeline: update resource name to uuid (instill-ai/controller#25) (97dd55d)
Console (v0.22.1-alpha)
Bug Fixes
- further fix test flakiness (instill-ai/console#412) (dcadf61)