Google Drive
Last updated
Last updated
Aissistant supports Google Doc and Google Sheet as assets
A lot of Aissistant users use Google Doc as Knowledge Base
Click “Assets” from the left sight panel of your Aissist dashboard, you can then build your knowledge base with Google Doc or Google Sheet.
Additionally, we offer an assets debugger on the same page. Here, you can enter keywords or questions, and the debugger will generate a list of relevant information along with its sources. This allows you to confirm the accuracy of the information retrieved. It also enables you to identify and resolve any discrepancies or conflicts within the returned data.
The doc/sheet provided to Aissistant needs to meet one of following requirements:
a) It is shared with support@aissistant.io
b) It is set to be “accessible to anyone with the link”
There are multiple reasons that we recommend using Google Doc/Sheet as a primary asset:
It’s free and widely used;
It is function complete as a knowledge base;
It is easier to manage (search, edit, index, etc);
It is easy for team collaboration;
Although the information in both places can play the same effect, they are significantly different.
Workspace Instruction: the content in workspace instruction is global which applies to all the interactions. This is where you put something globally important, for example “be brief on the response”, “never confirm the schedule time”, etc. However, workspace instruction is supposed to be short and brief, usually within 4K English characters (~ 1K English words). Otherwise, it won’t be very efficient to carry a big chunk of instruction on every interaction.
Google Doc: the content in Google Doc is not global therefore not guaranteed to be included in every interaction. This is where you put something that is contextually important, for example “the move-out procedure is …”, “our return policy is …”. It can be very long. However, it is very important to section the Google Doc by headings, because AI leverages the headings to intelligently select the relevant information to be included based on the context that conversation is in.
Another difference is that Workspace Instruction is meant to be more static, while Google Doc can be the place for dynamic content that has frequent updates. For more information, please refer to Fine-tune Aissistant Behavior.
You can included multiple Google Docs in one workspace.
We try to little restrictions on format, except the following two recommendations:
It is very important to use the HEADINGS to separate the topics in Google Doc. The heading can be a short description of topic or several example questions divided by ‘|’. Heading is an important manner to help AI segment your doc and intelligently select the relevant section to use.
Use the heading to separate the content, to help Aissistant better understand it. We usually recommend one section is no longer than 1 page.
Use “normal text” as the format of content which can be “paragraph”, “list” or even “table”
We don’t support image understanding in Google Doc yet.
We try to keep little restrictions on format, but with the following recommendations:
Keep each table small. Large table will make it challenging to effectively process. If needed, split the table by categories like product, customer, geographic location, etc. (This recommendation works for human user too)
Keep one table per tab. Multiple tables per tab sometimes confuse AI, and human too.
On the top of each table, have some description to explain the content and use of table below. It helps AI to understand, and human too.
Emphasize
Use [[ and ]] to enclose the text to instruct Aissistant to follow the provided content as much as possible in the response.
Contextual instruction
Use {{ and }} to enclose an instruction on how to handle this specific situation.
Be specific on the instruction, for example, include when, what, how and why.
The following assets are also supported, please contact sales@aissistant.io for assets that are not in the list.
Asset Type | Status |
---|---|
Google Doc
Released
Google Sheet
Released
Website
Released
Aissist Knowledge Base (embedded)
Released
SharePoint (Doc/Spreadsheet)
Beta
SQL Database
Beta
Restful API
Released
ETA: Q2 2024