api: adds AllowModeOverride for extproc#5099
Conversation
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
|
@arkodg @zirain @zhaohuabing ptal |
|
sorry i haven't added the test yet - just wanted to know if this is ok before then |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #5099 +/- ##
==========================================
- Coverage 66.87% 66.85% -0.03%
==========================================
Files 210 210
Lines 32953 32948 -5
==========================================
- Hits 22038 22027 -11
- Misses 9585 9589 +4
- Partials 1330 1332 +2 ☔ View full report in Codecov by Sentry. |
|
added (obvious) test |
|
+1 in general. Maybe this can be grouped in the "processing mode" container. e.g. |
|
hey @mathetake @guydc can you share the use case for this ? is it when the policy creator is unaware of what the processing mode should be and is set lazily by the ext proc service ? |
|
when dealing with OpenAI and any other AI chat completion endpoints, the returned content is conditionally streaming - the chat completion is sent line by line. so we are converting the response line by line and want to send them to the client as soon as possible not to block the giant entire events. |
|
thanks @mathetake this makes sense |
|
sounds good, will rework the pr then |
Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
|
@arkodg done |
|
no idea why the doc build is failing (seems irrelevant) |
|
maybe related to linkinator thats working hard to make sure link references are working |
* api: adds AllowModeOverride for extproc Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com> * add test Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com> * review: move it to inside processingMode Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com> --------- Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com> Signed-off-by: EshaanAgg <96648934+EshaanAgg@users.noreply.github.com>
This allows us to set allowModeOverride=true for our external processor. The flag is needed to deal with the stream=true requests. This is introduced to EG today: envoyproxy/gateway#5099 --------- Signed-off-by: Takeshi Yoneda <t.y.mathetake@gmail.com>
What type of PR is this?
Adds a new API for external processor
What this PR does / why we need it:
This adds AllowModeOverride boolean config
to the external processing config.
Which issue(s) this PR fixes:
N/A