You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
- "Grade suggestion is displayed and selected by default, but the instructor much make the final decision to edit or accept the suggestions."
12
+
+ "Grade suggestion is displayed and selected by default, but the instructor must make the final decision to edit or accept the suggestions."
13
+
```
14
+
15
+
### quickreassess
16
+
#### uid
17
+
```diff
18
+
+ "quickreassess"
19
+
```
20
+
#### revision
21
+
```diff
22
+
+ "2025.10.27"
23
+
```
24
+
#### feature
25
+
```diff
26
+
+ {
27
+
"name": "Suggestions for Quick Reassess",
28
+
"description": "Allows the use of AI to generate a question derived from the original. These questions can be re-generated, edited, and rejected or approved before they are published to students."
29
+
}
30
+
```
31
+
#### model
32
+
```diff
33
+
+ {
34
+
"name": "Claude Family Models",
35
+
"description": "",
36
+
"trained": "No",
37
+
"data": "Other",
38
+
"dataDescription": "The original question stem used to generate derivatives"
39
+
}
40
+
```
41
+
#### compliance
42
+
```diff
43
+
+ {
44
+
"retention": "Standard execution, infrastructure, and other operational logs are retained for a reasonable period to enable monitoring and troubleshooting of underlying services.",
45
+
"logging": "Does not log data",
46
+
"loggingDescription": "",
47
+
"regions": "Virginia",
48
+
"regionsDescription": "",
49
+
"pii": "Not Exposed",
50
+
"piiDescription": ""
51
+
}
52
+
```
53
+
#### outputs
54
+
```diff
55
+
+ {
56
+
"settings": "Yes",
57
+
"human": "Yes",
58
+
"humanDescription": "Educators have access to tools that allow them to preview, regenerate, edit, or replace AI generated questions before they are published in an assessment.",
59
+
"guardrails": "Only educators can generate items and all AI generated items are in draft status until reviewed and approved.",
60
+
"risks": "AI generated questions may require additional review and editing to ensure accuracy, relevance, and alignment with intended learning objectives.",
61
+
"outcomes": "Educators can generate new standards-aligned questions based on existing items in order to quickly create diverse, tailored assessments that expand beyond the current Mastery Item Bank content, enhancing instructional flexibility and efficiency."
Copy file name to clipboardExpand all lines: packages/aiinfo/cache/cache.csv
+3-1Lines changed: 3 additions & 1 deletion
Original file line number
Diff line number
Diff line change
@@ -2,10 +2,12 @@ canvasInboxTranslation,2025.09.10,Inbox Translation,"Translation of inbox messag
2
2
canvasCourseTranslation,2025.09.10,Discussions Translation,"Translation of Discussion threads (""Course AI Translation"" feature flag) across 10 languages.",Haiku 3,Anthropic Claude models are provided via Instructure's in-house AI Platform.,No,Course,Announcement and Discussion prompts and replies,Data is not stored or reused by the model.,Does not log data,,Global,,Not Exposed,"PII in discussion replies may be sent to the model, but no PII is intentionally sent to the model.",Yes,Yes,Untranslated content is available to review translations against,,Machine translation may not fully capture the meaning of the original message.,Improve participation for students who do not natively speak the language of instruction or other replies.,Canvas,2
3
3
IgniteAgent,2025.09.10,Agent,Ignite Agent is a faculty chat assistant capable of translating natural language requests into complex Canvas workflows.,"Haiku 3.5, Sonnet 3.5",Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs).,No,Other,The model uses chat transcripts and information requested from the Canvas API to execute its actions.,Chat logs are retained indefinitely for troubleshooting and debugging.,Logs data,Chat logs are retained for troubleshooting and debugging purposes.,"Virginia, Oregon",,Exposed,"If requested during the the chat, user, course, and student identifiers or metadata may be shared with the model.",Yes,Yes,"The agent only takes action based on human requests, and all write actions must be confirmed by the user.",Access to the agent is limited to users with an Admin- or Teacher-based role. Data access and functions are scoped to the permissions available to the chat user.,The model may misinterpret user requests and require additional prompting.,"Users are able to save time by relying on the Agent to execute complex workflows, batch actions, and other time-consuming Canvas tasks.",IgniteAI,2
4
4
rubricGenerator,2025.10.09,Rubric Creator,AI-generated rubric recommendations for Canvas assignments.,Haiku 3,Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs).,No,Course,The Canvas assignment description field is used in generating rubric criteria.,Model responses are stored for debugging purposes.,Logs data,"Request, response, and feedback data is logged to assist in troubleshooting.",Global,,Not Exposed,,Yes,Yes,All created rubrics are reviewable and editable by the instructor.,The instructor may edit criteria before accepting the rubric.,The created rubric may not align with the assignment's intended learning outcomes.,A rubric aligned to learning outcomes is created for the instructor to use during grading.,Canvas,2
5
-
canvasGradingAssistance,2025.10.22,Grading Assistance,"AI powered automatic grading and feedback suggestions to enable more consistent, fair, and efficient grading workflows.",Haiku 3,Anthropic Claude models are provided via Instructure's in-house AI Platform.,No,"Course, Student","Assignment information, rubric, and student submissions.",Transactional data is retained for the life of the request,Logs data,Complete response from the LLM is retained in the Canvas database for auditing purposes.,Global,,Not Exposed,"No PII is intentionally sent to the model. If there is incidental PII in any of the shared data, such as in the submission body, it will be sent to the model.",Yes,Yes,"Grade suggestion is displayed and selected by default, but the instructor much make the final decision to edit or accept the suggestions.",The Claude Haiku model has inherent guardrails built in.,Model may work better with certain kinds of assignment types or disciplines. The grader must remain alert to incorrect assumptions from the model.,"Reduced bias in grading, fairness, and consistency. Grading efficiency, saving teachers time.",Canvas,2
5
+
canvasGradingAssistance,2025.10.22,Grading Assistance,"AI powered automatic grading and feedback suggestions to enable more consistent, fair, and efficient grading workflows.",Haiku 3,Anthropic Claude models are provided via Instructure's in-house AI Platform.,No,"Course, Student","Assignment information, rubric, and student submissions.",Transactional data is retained for the life of the request,Logs data,Complete response from the LLM is retained in the Canvas database for auditing purposes.,Global,,Not Exposed,"No PII is intentionally sent to the model. If there is incidental PII in any of the shared data, such as in the submission body, it will be sent to the model.",Yes,Yes,"Grade suggestion is displayed and selected by default, but the instructor must make the final decision to edit or accept the suggestions.",The Claude Haiku model has inherent guardrails built in.,Model may work better with certain kinds of assignment types or disciplines. The grader must remain alert to incorrect assumptions from the model.,"Reduced bias in grading, fairness, and consistency. Grading efficiency, saving teachers time.",Canvas,2
6
6
discussionInsights,2025.10.02,Discussion Insights,"Discussion Insights uses AI to evaluate student discussion replies, highlight relevant contributions, and flag those that may need instructor review.",Haiku 3,Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs).,No,Course,"Discussion topic, prompt, and student replies are used.",No user data is stored or reused by the model.,Logs data,Model evaluations and reply labels are logged for debugging and troubleshooting purposes.,Global,,Exposed,"Known PII is masked before being sent to the model, though any PII present in the discussion reply is not and may be shared with the model.",Yes,Yes,Instructors may review AI-generated evaluations or review posts directly.,"Model responses are logged for quality assurance, and responses with low confidence are flagged ""Needs Review"" to encourage human intervention.",The model may misclassify some nuanced replies.,"Instructors are able to quickly assess the quality of student replies, identify low-effort or off-topic contributions, and focus their attention to where it is needed most.",Canvas,2
7
7
smartSearch,2025.10.02,Smart Search,Natural language search of course content for students and instructors.,Cohere Embed Multilingual,Cohere models are provided via Amazon Bedrock Foundation Models (FMs).,No,Course,Course content is indexed by the model and then stored in the Canvas database.,Data is not stored or reused by the model. Indexed course content is stored in the Canvas database.,Does not log data,,Global,,Not Exposed,"PII in course content may be indexed, but no PII is intentionally sent to the model.",Yes,Yes,Users are presented with a list of results related to their search query and can act or not act on them.,,Search results may be incorrectly sorted or may not be relevant to the search term.,"Students are able to quickly find answers to questions, and instructors are able to quickly navigate their courses.",Canvas,2
8
8
askyourdata,2025.09.12,Ask Your Data,AI-powered query tool enables users to ask natural language questions and receive textual or visual responses with detailed explanations of the methodology.,Doowii (multiple),"Doowii is a third-party sub-processor for Intelligent Insights. Doowii's tools use OpenAI GPT-4o, GPT-3.5 Turbo, and Claude Sonnet 3.5",No,Other,"Doowii is trained on the Canvas LMS database schema, and receives no data from Canvas. The user's prompt and heuristics (such as summary statistics) are shared with the model to generate a response.",Data is retained through the life of your contract with Instructure.,Logs data,,Global,,Exposed,"Prompt, summary statistics.",Yes,Yes,Ask your data returns a methodology description along with a generated query. Users have the ability to edit the generated SQL directly.,"Questions are scoped to the domain only, highly ambiguous terms ask for clarification.",Incorrect interpretation of the question. Inaccurate SQL may be generated. Suggested questions or methodologies may not be relevant.,Provide accurate data retrieval and analysis through natural language prompting.,Intelligent Insights,2
9
9
canvasDiscussionSummaries,2025.09.12,Discussion Summaries,"Discussion summaries of key points, questions, and insights.",Haiku 3,Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs).,No,Course,Discussion prompt and replies.,Data is not stored or reused by the model.,Logs data,Summaries are logged in the Canvas database for review and reuse.,Global,,Not Exposed,"Pii in discussion replies may be sent to the model, but no PII is intentionally sent to the model.",Yes,Yes,Instructors may regenerate summaries or fine-tune them with additional instructions.,,Summaries may not capture all desired information.,Summaries allow discussion topics to be used in large-format courses and other environments where discussions were previously infeasible.,Canvas,2
10
+
quickReassess,2025.10.27,Suggestions for Quick Reassess,"Allows the use of AI to generate a question derived from the original. These questions can be re-generated, edited, and rejected or approved before they are published to students.",Claude Family Models,,No,Other,The original question stem used to generate derivatives,"Standard execution, infrastructure, and other operational logs are retained for a reasonable period to enable monitoring and troubleshooting of underlying services.",Does not log data,,Virginia,,Not Exposed,,Yes,Yes,"Educators have access to tools that allow them to preview, regenerate, edit, or replace AI generated questions before they are published in an assessment.",Only educators can generate items and all AI generated items are in draft status until reviewed and approved.,"AI generated questions may require additional review and editing to ensure accuracy, relevance, and alignment with intended learning objectives.","Educators can generate new standards-aligned questions based on existing items in order to quickly create diverse, tailored assessments that expand beyond the current Mastery Item Bank content, enhancing instructional flexibility and efficiency.",Mastery,2
BCEAltText,2025.10.01,Block Content Editor Alt Text Generator,,Haiku 3,Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs).,No,Course,Content Editor Images.,Model responses are stored for debugging purposes.,Logs data,"Request, response, and feedback data is logged to assist in troubleshooting.",Global,,Not Exposed,,No,Yes,"User initiated, User must confirm output.",,Alt text might not always be accurate. ,An accessible Alt Text description of the image is generated.,Canvas,2
11
13
portfolios,2025.10.27,Portfolios,"AI Assistant for teachers to improve portfolio level and evidence level instructions improvement, skill extraction and reflection question generation.",Haiku 3,Anthropic Claude models are provided via Instructure's in-house AI Platform.,No,Other,Custom instructions that teachers provide in their Evaluation portfolios for students.,Data is not stored or re-used by the model.,Logs data,Model evaluations and reply labels are logged for debugging and troubleshooting purposes.,Global,,Not Exposed,,Yes,Yes,Teachers provide the input of the model (content) and the response by the model is not auto accepted. Teachers have the chance to modify the result every occasion.,"Teacher access to the model is limited, it's only through the instructions of the portfolio/evidence. There is suspicious prompt detection built in for extra security.","LLM might suggest not the right questions or skills, or not an improved instruction to the user's liking.","A user-friendly tool for a safe writing assistant, and features to speed up administrative work, and provide more reliable portfolios for students to work with.",Canvas,2
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?v=4&size=40){kind=avatar}
0 commit comments