A rival to OpenAI’s Deep Research Assistant Provide a built in AI agent that acts like a dedicated research assistant. You give it a prompt (a question or topic), and it: Searches the Web: It intelligently browses the internet, finding relevant information. Iterates: It doesn't just stop at the first search. It analyzes the results and asks itself, "Do I need more information?" This creates a loop of searching and refining. Generates a Report: Once it has gathered enough information, it compiles a complete report for you.

An Anonymous User about 1 year ago

27

New Model

💡

Feature Requests

Venice prides itself with being completely private, but users have no way of checking. They have to trust. I would appreciate a Privacy Audit done by a third party. Alternatively, making more of Venice’s Infrastructure public would increase trust in the Privacy claim. Is that possible / already done / in the works?

GitHappens about 1 year ago

27

API

💡

Feature Requests

The ability to create, effectively, a chat room with a human, and 2 custom characters where the AI’s will read and respond to all messages in turn. An important feature in such a mode would have to be a post order (or batting order for the Americans out there) so that the AI’s don’t end up bouncing off one another endlessly before the human can post. For example: Human posts Character 1 posts Character 2 posts -repeat- in this example, character 1 would not be able to post again until after the human as posted.

An Anonymous User about 1 year ago

27

Characters

💡

Feature Requests

As the title says, a model that is both uncensored and has vision. I think that would be my most requested feature.

An Anonymous User 12 months ago

20

New Model

+1

💡

Feature Requests

Currently, users must convert PDF files to text or read text files and manually include the content in the prompt when using the Venice API. This process is time-consuming, prone to errors, and inefficient, especially with large files. Proposed Solution: Introduce a feature allowing users to upload PDF or text files directly through the Venice API, similar to the functionality available in Venice Chat. This would eliminate the need for manual conversion or extraction, streamlining the workflow.

DMR2 about 1 year ago

5

API

API

Allow characters to proactively send messages without requiring a user prompt, enabling more interactive conversations.

Jack about 1 year ago

11

Characters

💡

Feature Requests

Venice has ability to analyze an uploaded image, and you have multiple image generation models, but they are only text to image, not text + reference image to image. All of the venice text to image models have difficulty positioning bodies precisely (such as in the schoolgirl pin wrestling position the wrestler is always in squat position with knees in air instead of against the mat and no level of prompting seems able to get the knees pressed down. A reference image would be very useful in such scenarios.

An Anonymous User 11 months ago

16

Image

💡

Feature Requests

Allow users to create custom distilled or fine-tuned models from base models users upload training data in some standardized format training progress can be monitored within the app after training, model is available for download and possibly use within Venice

An Anonymous User about 1 year ago

3

New Model

+2

💡

Feature Requests

**Goal:** Enable targeted AI edits to markdown files without overwriting manual changes, with version tracking. **Core Features:** - **Section-Specific AI Editing:** Use markers (e.g., `#section-id`) in markdown to let AI rewrite only specified parts (e.g., "Revise the 'Methods' section"). - **Version Control:** Auto-save snapshots after each edit (AI or manual), with diffs and rollback for any section. - **Efficiency:** Only regenerate targeted sections, preserving unchanged content. **Benefits:** - Faster iterations by editing fragments, not entire documents. - Audit changes with version history; revert/merge selectively. - Maintain formatting and manual edits alongside AI updates. **Example:** 1. Create a markdown file with `#abstract` and `#conclusion` sections. 2. Use an AI prompt to rewrite `#abstract` without affecting `#conclusion`. 3. View diffs and restore prior versions of specific sections. **Technical Notes:** - Store versions client-side (aligns with Venice’s privacy model). - Provide context to AI (e.g., surrounding text) for coherent section edits. **UI Needs:** - Sidebar for version history/diffs. - Inline "Edit with AI" buttons per section. **Impact:** Streamlines workflows for technical writing, documentation, and collaboration.

An Anonymous User 11 months ago

9

File Upload

💡

Feature Requests

this open source LLM is used for cyber security and pen testing. if venice integrates this, u can reach out to pentesting companies and security researchers to onboard to venice and VVV by unchainξd Post made by tame

An Anonymous User about 1 year ago

6

New Model

💡

Feature Requests

Ability to create projects with context files, like Claude projects. Currently we can abuse characters for this, but projects with option to save (multiple) files for context would be great for coding, for example.

arjan about 1 year ago

27

Chat

💡

Feature Requests

Have the ability to train a voice for your character. I think this would be a lot of fun and make the character more realistic. On this point it would be great if during a chat with the character, images are generated using the characters image and pose/elements are relevant to the current discussion.

An Anonymous User 12 months ago

10

Voice

+1

💡

Feature Requests

Create Lora powers. This means loading several photos of a character (usually around 50) to create the Lora. To be able to create images from them. And maybe even make it public to the community like RP characters.

An Anonymous User 11 months ago

6

Image

💡

Feature Requests

I might be missing something obvious here but why not have a desktop app that bypasses the issues and limitations of using a browser? Like not being able to use the max context window a model can offer and cache storage limits browsers have. Just seems like a desktop app would provide a far more standard experience than having to cater to multiple browsers as well as offer Venice more control in what features they can implement.

CountSix about 1 year ago

8

App

💡

Feature Requests

All browsers implement a standard called “OpenSearch”, which lets websites advertise that they support a search provider. Each browser implements the UX for this slightly differently, but all support it in some way or another. By adding a file to venice.ai, and adding a tag to the site, it’s possible for users to add Venice as a search provider to their browser. You can advertise multiple search options, such as “Venice: Llama 3.1 405B”, “Venice: Deepseek R1 671B”, “Venice: Default Web Model”. The UX from thereon out would be that you can set Venice as your default search provider, and searches would be sent to Venice instead of, say, Google or DDG. Here’s MDN’s reference material: https://developer.mozilla.org/en-US/docs/Web/XML/Guides/OpenSearch#reference_material Here’s what it looks like on IMDb when you right-click the search bar on Firefox, if you don’t already have IMDb as a search provider. Here’s what a search looks like.

Justin Martin about 1 year ago

2

Web Search

💡

Feature Requests

Incorporate two-way voice chat functionality on models. There are amazing fully open source models you can use now. Can that be done through web apps linked on the home screen?

An Anonymous User 11 months ago

4

Voice

💡

Feature Requests

https://en.wikipedia.org/wiki/Warrant_canary Venice should publish a warrant canary, covering a variety of specific claims. Here’s what Deepseek came up with: Venice Warrant Canary Last Updated: [Month/Day/Year] To our users: Venice LLC has not received any secret subpoenas, warrants, national security letters, gag orders, or other compelled surveillance mandates from any governmental entity that would compromise user privacy. No Compelled Disclosure We have never provided user data to any third party, including governments. We have not modified systems to enable surveillance. No Backdoors No cryptographic keys or system access have been surrendered to external actors. No Data Retention As a zero-knowledge platform, Venice cannot access or retain user prompts, documents, or inference outputs. If this statement is not updated within 30 days, assume we are legally barred from renewing it. Absence of this canary constitutes constructive notice of government coercion. Verification: Digitally signed with Venice’s root key (0x...). Historical versions archived via IPFS (CID: Qm...). Hash published on Base blockchain (VVV token contract). "We built Venice to resist tyranny. If forced to betray you, we will shut down before complying." — Venice Team Cryptographic Signature: -----BEGIN PGP SIGNATURE----- [LIVE-SIGNATURE-HERE] -----END PGP SIGNATURE----- (Note: Update timestamp and signature monthly. Host raw text at https://venice.ai/canary with immutable prior versions.)

Justin Martin about 1 year ago

1

Privacy

💡

Feature Requests

Russian, Spanish, German are the top feedback we have received thus far.

Gabe K 8 months ago

12

💡

Feature Requests

An option to download the chat history backup locally, and to import such a backup file. Backups stored on Venice servers can only be restored to the same account, this should allow importing to a different account. There are several reasons: To preserve chat history when merging separate accounts (e.g. email + web3 account). For people who don’t want to have their backup stored on Venice servers (despite the encryption). There is a related request, but I think this is a separate feature: https://veniceai.featurebase.app/p/batch-export-chats For using the data outside the web app. If it could also be imported into the web app, then it would serve the same purpose.

arjan about 1 year ago

7

Backup / Sync

💡

Feature Requests

On non trivial, long terme discussion, particularly a serious RPG context, the character have a tendency to forget their state if this state is not repeated in every message. As an exemple, a fighter can lost his sword, go for a long dialog, some other action, and at the next fight, it will draws his sword (the one he didn’t have anymore!). That occurs because not all history is included with each interaction (make sense) and therefore the message where he lost his weapon can be lost. The more the discussion goes, the worst it become in a complexe history. The AI can start to forget whole part of the discussion, or even his current goal, which break the immersion. This could be fixed with a state file. Each interaction response could contain 2 parts: the message displayed in the conversation, and an updated state which is updated locally. The state file could contain information about the main character, user persona, other secondary characters, world/environnement etc. I believe that the state file should be viewable and editable, to allow fixing errors.

An Anonymous User about 1 year ago

6

Chat

💡

Feature Requests