Replies: 2 comments
-
Thanks for the screenshots. What do you mean by "Sources will flood the chat with results if I add that many"? Are you talking about the display of Sources being too long? The sources isn't ranked by relevance in Vault QA (basic), unlike in Plus. If a collapsed "Sources" section solves your problem, let me know. Setting max sources to 10 sounds pretty reasonable and I can make it as default too. (And no need to have |
Beta Was this translation helpful? Give feedback.
-
I think overall I'm not clear how this works or how the plus features work. I think I only recently noticed this because I was trying to proactively use copilot to help plan my day or catch me up on articles I wanted to read. Features outside of locating those items have been working phenomenally for me. My workflow before was pretty much almost all /copilot commands (or right click) in a single note. Multiple notes are giving me challenges. Maybe a future feature that would also solve this might be to have a @[folder name] like I may be doing something very wrong and/or have a bad expectation. I will also mention that my Obsidian Vault(s) is very small and I have never had one of those massive millions-of-notes type of Vaults. The workflow I want and the expectation I have may not align with users that have massive vaults. Even more context The example I filed the bug/discussion about has to do with feeding a pet. I also have TODOs/Notes for things like taking out the recycling - which happens every 3 days. If I query when I last took out the recycling I'll get a note from 2023 (or some other "random-ish" date) even though I did it yesterday. This isn't helpful. I mentioned Obsidian Webclippings and the problem I have there is podcasts. I listen to a podcast with weekly releases and I often use the Web Clippings to capture the podcast and go back to listen and make notes later. If I ask Copilot about topics that repeat in the podcast I will again get something from 2023 (or random-ish date in the middle). Sorry for the wall of text but I am struggling to wrap my head around how to explain what I am trying to do while simultaneously questioning if I'm trying to do something crazy. Inreases to Max Sources and a collapsible Sources section would solve the problem in the short term but I feel like that would not scale for vaults with > 100 notes of repeated common tasks/topics. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
I have a vault with Daily Notes and Tasks plugins. Each day ex: 2025-01-26 there may be a task ie:
- [x] #task #animalfeeding Fed [[Toffee]]
. When I ask "When was the last time I fed toffee?" with the default Max Sources of 3 the chat will return something from last year. If I increase the results to 10 it will find the latest one. The problem being the Sources will flood the chat with results if I add that many.I have tried this with multiple models and embedding models. The one I use the most is Gemini 1.5-pro with Gemini embeddings. I have also tried gemini flash, deepseek-v2, mistral, llama + nomic-embed-text and the results are the same.
I tried this same experiment with another plugin that by default seems to send 100 documents from Obsidian at a time and the answer was correct.
I thought this was a bug but maybe I have something misconfigured.
Screenshot attached. The only difference between these two queries is that the first one has max sources set to 3, second one has Max Sources set to 50.

This is a screenshto of the Smart Second Brain plugin with the same LLM configuration.

What am I doing wrong with Copilot?
Beta Was this translation helpful? Give feedback.
All reactions