new feature flag for situation based latency #1202
Conversation
|
@philipp-spiess updated based on your feedbacks (thank you!) |
vscode/src/completions/latency.ts
Outdated
| } | ||
|
|
||
| if (!lastCandidate?.result.logId) { | ||
| user = currentUserLatency ? currentUserLatency * 2 : defaultLatency.user |
There was a problem hiding this comment.
I think we need to set the new value to currentUserLatency right? Otherwise this will always be 0.
I think we should also make sure this can't grow unbounded so maybe we do
user = currentUserLatency = currentUserLatency ? Math.min(currentUserLatency * 2, defaultLatency.maxLatency) : defaultLatency.user
Co-authored-by: Philipp Spiess <hello@philippspiess.com>
| expect(getLatency(provider, languageId)).toBe(1400) | ||
| }) | ||
|
|
||
| it('returns increasing latency up to max latency for supported language on non-anthropic provider after rejecting multiple suggestions consistently', () => { |
There was a problem hiding this comment.
These tests are pure 🔥
There was a problem hiding this comment.
Agree, great work @abeatrix 🔥🔥🔥
There was a problem hiding this comment.
@philipp-spiess thank you so much for your help, testing is hardddd 😭
@arafatkatze thank you! I'm sorry we haven't had time to review your PRs. Our team is currently busy working on the release next week, so we apologize for being unresponsive lately, but we want to let you know that we appreciate your help and your feedback in the past few weeks, and we should be able to catch up on your works after the release 🙇♀️
There was a problem hiding this comment.
@abeatrix No worries at all! I completely understand the crunch you're in with the upcoming release. I'm genuinely impressed by the creativity that goes behind all of this work and I couldn't help myself from taking a look and adding some thoughts. I'll keep exploring the codebase and trying to understand/implement some ideas. Don't feel pressured to review my PRs—I know you folks are busy, and they can totally wait until you have the bandwidth.
Just happy to help in any way I can!!
CLOSE: #1189, #1175, #1176, #1177 feat: new feature flag for situation-based latency Add new feature flag for autocomplete to adjust minimum latency for completions based on env and user actions - Remove separate feature flags for 350ms and 600ms minimum latency - Add single feature flag for adjustable minimum latency - Introduce LatencyManager to adjust latency based on user actions - Apply adjusted minimum latency before showing completions <!-- Required. See https://docs.sourcegraph.com/dev/background-information/testing_principles. --> Expected behavior: 1. every user starts at `200ms` under `user latency` 2. `baseline latency` starts at `400ms` for all non-anthropic provider, `0ms` for anthropic 3. working in low performances files adds additional `1200ms` to the `baseline latency` under `additional latency` 4. when a suggestion is accepted, `baseline latency` reset to starting `baseline latency`, `user latency` reset to 0 5. when a suggestion is not accepted (e.g. typing over ghost type), double the current `user latency` count. 6. typing over the same ghost type counts as one reject only, and will not increase latency again https://github.com/sourcegraph/cody/assets/68532117/9f287dcf-fb60-435b-b37e-f7446fad5cf3 --------- Co-authored-by: Philipp Spiess <hello@philippspiess.com>
CLOSE: #1189, #1175, #1176, #1177
feat: new feature flag for situation-based latency
Add new feature flag for autocomplete to adjust minimum latency for completions based on env and user actions
Test plan
Expected behavior:
200msunderuser latencybaseline latencystarts at400msfor all non-anthropic provider,0msfor anthropic1200msto thebaseline latencyunderadditional latencybaseline latencyreset to startingbaseline latency,user latencyreset to 0user latencycount.add.latency.for.throttling.autocompletes.in.unwanted.and.low.performances.environments.mp4