fix(proxy): forward Bedrock event-stream content-type on unbuffered passthrough by devin-ai-integration[bot] · Pull Request #33767 · BerriAI/litellm

devin-ai-integration · 2026-07-17T19:49:17Z

TLDR

Problem this solves:

Bedrock passthrough streaming responses carried no content-type header
Claude Code past 2.1.207 rejects such streams as gateway-corrupted
Newer versions silently refetch non-streaming, doubling cost and latency

How it solves it:

The unbuffered passthrough stream now sets media_type from the provider handler
Same helper the buffered guardrail branch already passes
Bedrock streams return application/vnd.amazon.eventstream
Providers with no registered event-stream media type stay unchanged

User Flow

Before: a developer running Claude Code against the proxy's Bedrock passthrough sees every streaming turn rejected or silently doubled

They set CLAUDE_CODE_USE_BEDROCK=1 and ANTHROPIC_BEDROCK_BASE_URL=https://litellm-domain/bedrock, then ask Claude Code a question
Claude Code sends POST https://litellm-domain/bedrock/model/{model}/invoke-with-response-stream
The response streams valid AWS event-stream bytes but carries no content-type header
Claude Code errors with API Error: Bedrock streaming response has content-type "application/octet-stream"; expected "application/vnd.amazon.eventstream", and versions past 2.1.227 instead silently refetch the same turn with stream: false, so every turn costs two full model invocations

After: the same turn streams once and renders normally

They set CLAUDE_CODE_USE_BEDROCK=1 and ANTHROPIC_BEDROCK_BASE_URL=https://litellm-domain/bedrock, then ask Claude Code a question
Claude Code sends POST https://litellm-domain/bedrock/model/{model}/invoke-with-response-stream
The response streams the same bytes with content-type: application/vnd.amazon.eventstream
Claude Code renders the streamed answer in one pass, with no guard error and no duplicate non-streaming call

Relevant issues

Linear ticket

Resolves LIT-4561
Resolves LIT-5674

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

I have added meaningful tests
My PR passes all CI/CD checks (e.g., lint, format, unit tests)
My PR's scope is as isolated as possible; it only solves 1 specific problem
I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

Screenshots / Proof of Fix

Two live proxies against real AWS Bedrock (bedrock/us.anthropic.claude-opus-5, us-east-1, real spend), each booted from its own worktree and venv with its own fresh Postgres, model_list router deployment us.anthropic.claude-opus-5 (custom_llm_provider: bedrock), no mocks. Before runs the merge base f04b1da on port 53571, After runs this PR's tip d5a4c14 on port 33945. Same payloads both sides. The Claude Code case runs the real claude CLI 2.1.234 interactively under tmux with CLAUDE_CODE_USE_BEDROCK=1, ANTHROPIC_BEDROCK_BASE_URL=http://127.0.0.1:{port}/bedrock, CLAUDE_CODE_SKIP_BEDROCK_AUTH=1, ANTHROPIC_AUTH_TOKEN=$VKEY (a per-leg virtual key from POST /key/generate), ANTHROPIC_MODEL=us.anthropic.claude-opus-5, and the same prompt, and the double invocation shows up through the end-user spend surfaces (GET /spend/logs?api_key=$VKEY, GET /key/info?key=$VKEY)

Before (`f04b1da`)

invoke-with-response-stream

Customer's header check

$ curl -s -D headers.txt -o body.bin -X POST http://127.0.0.1:53571/bedrock/model/us.anthropic.claude-opus-5/invoke-with-response-stream \
    -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" \
    -d '{"anthropic_version":"bedrock-2023-05-31","max_tokens":32,"messages":[{"role":"user","content":"Say hi in three words"}]}'
$ grep -i '^HTTP\|^content-type\|^x-litellm-model-name' headers.txt
HTTP/1.1 200 OK
x-litellm-model-name: bedrock/us.anthropic.claude-opus-5
$ grep -ci '^content-type' headers.txt
0
$ xxd body.bin | head -2
00000000: 0000 02f4 0000 004b 7450 10e7 0b3a 6576  .......KtP...:ev
00000010: 656e 742d 7479 7065 0700 0563 6875 6e6b  ent-type...chunk

Streamed AWS event-stream bytes arrive with no content-type header at all, which is what makes Claude Code treat the stream as corrupted

converse-stream

Same check on the Converse streaming route

$ curl -s -D headers.txt -o body.bin -X POST http://127.0.0.1:53571/bedrock/model/us.anthropic.claude-opus-5/converse-stream \
    -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" \
    -d '{"messages":[{"role":"user","content":[{"text":"Say hi in three words"}]}],"inferenceConfig":{"maxTokens":32}}'
$ grep -i '^HTTP\|^content-type\|^x-litellm-model-name' headers.txt
HTTP/1.1 200 OK
x-litellm-model-name: bedrock/us.anthropic.claude-opus-5
$ grep -ci '^content-type' headers.txt
0
$ xxd body.bin | head -2
00000000: 0000 00b0 0000 0052 f080 e7a5 0b3a 6576  .......R.....:ev
00000010: 656e 742d 7479 7065 0700 0c6d 6573 7361  ent-type...messa

No content-type header on the Converse stream either

invoke (non-streaming control)

Non-streaming route, unchanged by this PR

$ curl -s -D headers.txt -o body.json -X POST http://127.0.0.1:53571/bedrock/model/us.anthropic.claude-opus-5/invoke \
    -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" \
    -d '{"anthropic_version":"bedrock-2023-05-31","max_tokens":32,"messages":[{"role":"user","content":"Say hi in three words"}]}'
$ grep -i '^HTTP\|^content-type\|^x-litellm-model-name' headers.txt
HTTP/1.1 200 OK
content-type: application/json
x-litellm-model-name: bedrock/us.anthropic.claude-opus-5
$ head -c 200 body.json
{"model":"claude-opus-5","id":"msg_bdrk_mqtio6tzsztbnpdh2pmzqn7og542inwylmxvkfqlllftjrcwnara","type":"message","role":"assistant","content":[{"type":"text","text":"Hi there, friend!"}],"stop_reason":"

content-type: application/json on both sides, so the non-streaming path is untouched

Claude Code 2.1.234 turn (Bedrock gateway mode)

Mint a virtual key for the session: curl -s -X POST http://127.0.0.1:53571/key/generate -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" -d '{"key_alias":"claude-code-before"}' and export its key as $VKEY
Launch the CLI interactively in tmux: CLAUDE_CODE_USE_BEDROCK=1 ANTHROPIC_BEDROCK_BASE_URL=http://127.0.0.1:53571/bedrock CLAUDE_CODE_SKIP_BEDROCK_AUTH=1 ANTHROPIC_AUTH_TOKEN=$VKEY ANTHROPIC_MODEL=us.anthropic.claude-opus-5 ANTHROPIC_SMALL_FAST_MODEL=us.anthropic.claude-opus-5 ANTHROPIC_DEFAULT_HAIKU_MODEL=us.anthropic.claude-opus-5 CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 claude --model us.anthropic.claude-opus-5, accept the folder trust prompt, type reply with exactly: STREAMING OK BEFORE and press Enter
Pane capture (tmux capture-pane -p)

│   Opus 5 with xhigh effort · Amazon Bedrock   │
❯ reply with exactly: STREAMING OK BEFORE
⏺ STREAMING OK BEFORE
✻ Cooked for 5s

What the proxy billed that one turn to the key, through the end-user spend routes

$ curl -s "http://127.0.0.1:53571/spend/logs?api_key=$VKEY" -H "Authorization: Bearer $LITELLM_KEY" \
    | jq '[.[] | select(.status=="success")] | map({call_type, model, prompt_tokens, completion_tokens, spend})'
[
 {
  "call_type": "allm_passthrough_route",
  "model": "bedrock/us.anthropic.claude-opus-5",
  "prompt_tokens": 31076,
  "completion_tokens": 14,
  "spend": 0.0174867
 },
 {
  "call_type": "allm_passthrough_route",
  "model": "bedrock/us.anthropic.claude-opus-5",
  "prompt_tokens": 993,
  "completion_tokens": 28,
  "spend": 0.0062315
 },
 {
  "call_type": "allm_passthrough_route",
  "model": "bedrock/us.anthropic.claude-opus-5",
  "prompt_tokens": 31076,
  "completion_tokens": 14,
  "spend": 0.21402975
 },
 {
  "call_type": "allm_passthrough_route",
  "model": "bedrock/us.anthropic.claude-opus-5",
  "prompt_tokens": 993,
  "completion_tokens": 16,
  "spend": 0.0059015
 }
]
$ curl -s "http://127.0.0.1:53571/key/info?key=$VKEY" -H "Authorization: Bearer $LITELLM_KEY" | jq '{key_alias: .info.key_alias, spend: .info.spend}'
{
 "key_alias": "claude-code-before",
 "spend": 0.24364945
}

Claude Code 2.1.234 renders the answer but only because it silently refetched: the single turn produced two ~31k-token main invocations and two ~1k-token side calls (each request run once as the rejected stream and once again as stream: false), and the key was billed for all four

After (`d5a4c14`)

invoke-with-response-stream

Customer's header check

$ curl -s -D headers.txt -o body.bin -X POST http://127.0.0.1:33945/bedrock/model/us.anthropic.claude-opus-5/invoke-with-response-stream \
    -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" \
    -d '{"anthropic_version":"bedrock-2023-05-31","max_tokens":32,"messages":[{"role":"user","content":"Say hi in three words"}]}'
$ grep -i '^HTTP\|^content-type\|^x-litellm-model-name' headers.txt
HTTP/1.1 200 OK
x-litellm-model-name: bedrock/us.anthropic.claude-opus-5
content-type: application/vnd.amazon.eventstream
$ grep -ci '^content-type' headers.txt
1
$ xxd body.bin | head -2
00000000: 0000 02d3 0000 004b 07b1 e3f3 0b3a 6576  .......K.....:ev
00000010: 656e 742d 7479 7065 0700 0563 6875 6e6b  ent-type...chunk

The same event-stream bytes now arrive with content-type: application/vnd.amazon.eventstream

converse-stream

Same check on the Converse streaming route

$ curl -s -D headers.txt -o body.bin -X POST http://127.0.0.1:33945/bedrock/model/us.anthropic.claude-opus-5/converse-stream \
    -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" \
    -d '{"messages":[{"role":"user","content":[{"text":"Say hi in three words"}]}],"inferenceConfig":{"maxTokens":32}}'
$ grep -i '^HTTP\|^content-type\|^x-litellm-model-name' headers.txt
HTTP/1.1 200 OK
x-litellm-model-name: bedrock/us.anthropic.claude-opus-5
content-type: application/vnd.amazon.eventstream
$ grep -ci '^content-type' headers.txt
1
$ xxd body.bin | head -2
00000000: 0000 0083 0000 0052 1601 25f3 0b3a 6576  .......R..%..:ev
00000010: 656e 742d 7479 7065 0700 0c6d 6573 7361  ent-type...messa

The Converse stream carries content-type: application/vnd.amazon.eventstream too, so the fix is not tied to one Bedrock streaming route

invoke (non-streaming control)

Non-streaming route, unchanged by this PR

$ curl -s -D headers.txt -o body.json -X POST http://127.0.0.1:33945/bedrock/model/us.anthropic.claude-opus-5/invoke \
    -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" \
    -d '{"anthropic_version":"bedrock-2023-05-31","max_tokens":32,"messages":[{"role":"user","content":"Say hi in three words"}]}'
$ grep -i '^HTTP\|^content-type\|^x-litellm-model-name' headers.txt
HTTP/1.1 200 OK
content-type: application/json
x-litellm-model-name: bedrock/us.anthropic.claude-opus-5
$ head -c 200 body.json
{"model":"claude-opus-5","id":"msg_bdrk_nx2bcw7z5cdwcecgvmfxgrfb23blyg2usfzovba6f4rz6uds24hq","type":"message","role":"assistant","content":[{"type":"text","text":"Hi there, friend!"}],"stop_reason":"

content-type: application/json on both sides, so the non-streaming path is untouched

Claude Code 2.1.234 turn (Bedrock gateway mode)

Mint a virtual key for the session: curl -s -X POST http://127.0.0.1:33945/key/generate -H "Authorization: Bearer $LITELLM_KEY" -H "Content-Type: application/json" -d '{"key_alias":"claude-code-after"}' and export its key as $VKEY
Launch the CLI interactively in tmux: CLAUDE_CODE_USE_BEDROCK=1 ANTHROPIC_BEDROCK_BASE_URL=http://127.0.0.1:33945/bedrock CLAUDE_CODE_SKIP_BEDROCK_AUTH=1 ANTHROPIC_AUTH_TOKEN=$VKEY ANTHROPIC_MODEL=us.anthropic.claude-opus-5 ANTHROPIC_SMALL_FAST_MODEL=us.anthropic.claude-opus-5 ANTHROPIC_DEFAULT_HAIKU_MODEL=us.anthropic.claude-opus-5 CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 claude --model us.anthropic.claude-opus-5, accept the folder trust prompt, type reply with exactly: STREAMING OK AFTER and press Enter
Pane capture (tmux capture-pane -p)

│   Opus 5 with xhigh effort · Amazon Bedrock   │
❯ reply with exactly: STREAMING OK AFTER
⏺ STREAMING OK AFTER
✻ Worked for 3s

What the proxy billed that one turn to the key, through the end-user spend routes

$ curl -s "http://127.0.0.1:33945/spend/logs?api_key=$VKEY" -H "Authorization: Bearer $LITELLM_KEY" \
    | jq '[.[] | select(.status=="success")] | map({call_type, model, prompt_tokens, completion_tokens, spend})'
[
 {
  "call_type": "allm_passthrough_route",
  "model": "bedrock/us.anthropic.claude-opus-5",
  "prompt_tokens": 31075,
  "completion_tokens": 13,
  "spend": 0.118184275
 },
 {
  "call_type": "allm_passthrough_route",
  "model": "bedrock/us.anthropic.claude-opus-5",
  "prompt_tokens": 992,
  "completion_tokens": 27,
  "spend": 0.0061985
 }
]
$ curl -s "http://127.0.0.1:33945/key/info?key=$VKEY" -H "Authorization: Bearer $LITELLM_KEY" | jq '{key_alias: .info.key_alias, spend: .info.spend}'
{
 "key_alias": "claude-code-after",
 "spend": 0.124382775
}

The same turn produces exactly one ~31k-token main invocation and one ~1k-token side call, and the key is billed once per request; the streamed answer renders in one pass with no guard error and no non-streaming refetch

Observations from the run (this PR leaves them alone):

Both legs: startup model-discovery probes fail, models absent from model_list
Passthrough x-litellm-response-cost* headers read 0.0, spend logs correct

Type

🐛 Bug Fix

Caveats (if any)

Final Attestation

The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

Link to Devin session: https://app.devin.ai/sessions/835c506325e2454d80dc4434ae74f7b3
Requested by: @shivamrawat1

d5a4c14 passes /live-pr-risk

…assthrough Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

devin-ai-integration · 2026-07-17T19:49:21Z

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

Disable automatic comment, CI, and merge conflict monitoring

greptile-apps · 2026-07-17T19:51:17Z

Greptile Summary

This PR fixes a missing Content-Type header on the unbuffered Bedrock passthrough streaming path by forwarding the provider's event-stream media type on the StreamingResponse, exactly as the buffered guardrail branch already did. The change is a single-line addition that reuses the existing _passthrough_event_stream_media_type() helper, keeping the fix provider-agnostic.

The one-liner media_type=self._passthrough_event_stream_media_type() is added to the unbuffered StreamingResponse (line 2042), matching the call already present on the buffered Response branch (line 2033).
Two focused mock-only regression tests are added: one pinning application/vnd.amazon.eventstream for Bedrock, and one confirming non-Bedrock providers continue to emit no content-type header.

Confidence Score: 5/5

Safe to merge — the change is a one-liner that applies an existing helper to a second code path, backed by concrete before/after evidence and passing mock tests.
The fix is minimal and surgical: one helper call added to the unbuffered stream path to match behaviour that was already correct on the buffered path. No logic is changed, no new abstractions are introduced, and the existing _passthrough_event_stream_media_type() already had test coverage for the buffered branch. The two new tests cover both the positive (Bedrock gets the correct header) and negative (non-Bedrock is unchanged) cases without real network calls, and the PR description includes live curl output confirming the fix against real AWS Bedrock endpoints.
No files require special attention.

Important Files Changed

Filename	Overview
litellm/proxy/common_request_processing.py	Adds `media_type=self._passthrough_event_stream_media_type()` to the unbuffered `StreamingResponse` branch, mirroring the already-existing call in the buffered `Response` branch. Change is minimal and correct.
tests/test_litellm/proxy/test_common_request_processing.py	Adds two new mock-only tests: one pinning the Bedrock unbuffered stream to `application/vnd.amazon.eventstream`, and one confirming non-Bedrock providers emit no content-type header. No real network calls; assertions are precise.

_{Reviews (3): Last reviewed commit: "docs(proxy): pre-fix passthrough streams..." | Re-trigger Greptile}

codecov · 2026-07-17T19:55:43Z

Codecov Report

❌ Patch coverage is 0% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
litellm/proxy/common_request_processing.py	0.00%	4 Missing ⚠️

📢 Thoughts on this report? Let us know!

codspeed-hq · 2026-07-17T20:10:27Z

Merging this PR will not alter performance

✅ 31 untouched benchmarks

_{Comparing litellm_lit_4561_bedrock_passthrough_content_type (d5a4c14) with litellm_internal_staging (2bc87ec)¹}

No successful run was found on litellm_internal_staging (cde5465) during the generation of this report, so 2bc87ec was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

devin-ai-integration · 2026-07-17T20:45:13Z

Before/after proof against two live proxies hitting real AWS Bedrock (global.anthropic.claude-opus-4-8), same invoke-with-response-stream request to each. Port 4001 runs the base branch (before), port 4000 runs this PR (after)

BEFORE (base branch, port 4001)

$ curl -s -D - -o /tmp/before.bin -X POST \
  "http://localhost:4001/bedrock/model/global.anthropic.claude-opus-4-8/invoke-with-response-stream" \
  -H "Authorization: Bearer sk-1234" -H "Content-Type: application/json" \
  -d '{"anthropic_version":"bedrock-2023-05-31","max_tokens":30,"messages":[{"role":"user","content":"say hello in 3 words"}]}' | grep -iE '^(HTTP|content-type)'
HTTP/1.1 200 OK

No content-type header, which is what trips the Claude Code >2.1.207 guard

AFTER (this PR, port 4000)

$ curl -s -D - -o /tmp/after.bin -X POST \
  "http://localhost:4000/bedrock/model/global.anthropic.claude-opus-4-8/invoke-with-response-stream" \
  -H "Authorization: Bearer sk-1234" -H "Content-Type: application/json" \
  -d '{"anthropic_version":"bedrock-2023-05-31","max_tokens":30,"messages":[{"role":"user","content":"say hello in 3 words"}]}' | grep -iE '^(HTTP|content-type)'
HTTP/1.1 200 OK
content-type: application/vnd.amazon.eventstream

Body is still the raw AWS event-stream, unchanged by setting the header

$ xxd /tmp/after.bin | head -4
00000000: 0000 02e9 0000 004b ec20 43d4 0b3a 6576  .......K. C..:ev
00000010: 656e 742d 7479 7065 0700 0563 6875 6e6b  ent-type...chunk
00000020: 0d3a 636f 6e74 656e 742d 7479 7065 0700  .:content-type..
00000030: 1061 7070 6c69 6361 7469 6f6e 2f6a 736f  .application/jso

Session: https://app.devin.ai/sessions/835c506325e2454d80dc4434ae74f7b3

…itellm_lit_4561_bedrock_passthrough_content_type

…ype header

devin-ai-integration · 2026-08-17T20:14:22Z

❌ Cannot revive Devin session - the session is too old. Please start a new session instead.

View session

mateo-berri · 2026-08-17T20:14:27Z

@greptileai

…tet-stream

mateo-berri · 2026-08-17T20:21:51Z

@greptileai

mateo-berri · 2026-08-18T20:12:38Z

bugbot run

cursor

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit d5a4c14. Configure here.}

mateo-berri

LGTM

fix(proxy): forward Bedrock event-stream content-type on unbuffered p…

371fa67

…assthrough Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

shivamrawat1 self-assigned this Jul 17, 2026

greptile-apps Bot reviewed Jul 17, 2026

View reviewed changes

Comment thread tests/test_litellm/proxy/test_common_request_processing.py Outdated

yassin-berriai mentioned this pull request Jul 17, 2026

refactor(e2e): remove bob_the_builder; drive remediation from a Grafana alert (provisioned outside the repo) #33749

Merged

4 tasks

mateo-berri added 2 commits August 17, 2026 13:04

Merge remote-tracking branch 'origin/litellm_internal_staging' into l…

8ced9f5

…itellm_lit_4561_bedrock_passthrough_content_type

test(proxy): assert non-Bedrock passthrough stream emits no content-t…

f86dc8f

…ype header

docs(proxy): pre-fix passthrough streams omitted content-type, not oc…

d5a4c14

…tet-stream

cursor Bot reviewed Aug 18, 2026

View reviewed changes

mateo-berri approved these changes Aug 18, 2026

View reviewed changes

mateo-berri merged commit 035bd76 into litellm_internal_staging Aug 18, 2026
75 of 76 checks passed

mateo-berri deleted the litellm_lit_4561_bedrock_passthrough_content_type branch August 18, 2026 20:43

Navigation Menu

Uh oh!

Uh oh!

fix(proxy): forward Bedrock event-stream content-type on unbuffered passthrough - #33767

fix(proxy): forward Bedrock event-stream content-type on unbuffered passthrough#33767
mateo-berri merged 4 commits into
litellm_internal_stagingfrom
litellm_lit_4561_bedrock_passthrough_content_type

devin-ai-integration Bot commented Jul 17, 2026 •

edited by mateo-berri

Loading

Uh oh!

devin-ai-integration Bot commented Jul 17, 2026

Uh oh!

greptile-apps Bot commented Jul 17, 2026 •

edited

Loading

Important Files Changed

Uh oh!

Uh oh!

codecov Bot commented Jul 17, 2026 •

edited

Loading

Uh oh!

codspeed-hq Bot commented Jul 17, 2026 •

edited

Loading

Uh oh!

devin-ai-integration Bot commented Jul 17, 2026

Uh oh!

devin-ai-integration Bot commented Aug 17, 2026

Uh oh!

mateo-berri commented Aug 17, 2026

Uh oh!

mateo-berri commented Aug 17, 2026

Uh oh!

mateo-berri commented Aug 18, 2026

Uh oh!

cursor Bot left a comment

Uh oh!

mateo-berri left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

devin-ai-integration Bot commented Jul 17, 2026 • edited by mateo-berri Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

TLDR

User Flow

Relevant issues

Linear ticket

Pre-Submission checklist

Delays in PR merge?

Screenshots / Proof of Fix

Before (f04b1da)

invoke-with-response-stream

converse-stream

invoke (non-streaming control)

Claude Code 2.1.234 turn (Bedrock gateway mode)

After (d5a4c14)

invoke-with-response-stream

converse-stream

invoke (non-streaming control)

Claude Code 2.1.234 turn (Bedrock gateway mode)

Type

Caveats (if any)

Final Attestation

Uh oh!

devin-ai-integration Bot commented Jul 17, 2026

🤖 Devin AI Engineer

Uh oh!

greptile-apps Bot commented Jul 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Uh oh!

Uh oh!

codecov Bot commented Jul 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

codspeed-hq Bot commented Jul 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will not alter performance

Footnotes

Uh oh!

devin-ai-integration Bot commented Jul 17, 2026

Uh oh!

devin-ai-integration Bot commented Aug 17, 2026

Uh oh!

mateo-berri commented Aug 17, 2026

Uh oh!

mateo-berri commented Aug 17, 2026

Uh oh!

mateo-berri commented Aug 18, 2026

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

mateo-berri left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

devin-ai-integration Bot commented Jul 17, 2026 •

edited by mateo-berri

Loading

Before (`f04b1da`)

After (`d5a4c14`)

greptile-apps Bot commented Jul 17, 2026 •

edited

Loading

codecov Bot commented Jul 17, 2026 •

edited

Loading

codspeed-hq Bot commented Jul 17, 2026 •

edited

Loading