Lot's of agent stuff again by satti-hari-krishna-reddy · Pull Request #343 · Shuffle/shuffle-shared

satti-hari-krishna-reddy · 2026-03-12T14:25:11Z

Here are the key changes:

Changed abort status from "FINISHED" to "ABORTED" throughout the codebase for clearer execution state tracking
Moved abort logic into the abortAgentExecution() function, replacing repetitive error handling code in ~7 locations
Fixed Action Reference bug where the wrong action object was being passed to the agent continuation logic. Previously, it was using actionResult.Action, but now it properly retrieves the original action from the execution results array, ensuring the agent has the correct state and context when continuing execution. (Look at line 17846 in shared.go)
Added context (ctx) as a parameter to HandleAiAgentExecutionStart()
Added validation for caller function info with trace ID logging for better debugging
Added "http" into the list of available apps for agent decisions), making the agent work with HTTP functionality again.
Added AI_AGENT_LLM_FAILURE log prefix for all LLM-related errors
Includes detailed error context: org ID, status codes, error types, and raw responses
Helps distinguish between different failure modes (auth errors, rate limits, parsing errors, etc.)
Added caller function and trace ID logging for agent starts
Better separation between ABORTED, FAILURE, and FINISHED states

frikky · 2026-03-12T21:34:28Z

+	formattedAppName := strings.ReplaceAll(strings.ToLower(app.Name), " ", "_")
+
+	isInternalShuffleApp := false
+	switch formattedAppName {


Something about just matching name seems off, but I'm not sure exactly why.

I wonder if there is a way to "steal" the URL field/apikey by formatting an app in a certain way.

Just need to think it over before merging 🤔

Yeah, I think having a special verified field for apps built by support users would be better. Given that we're already overriding the URL field when injecting the auth, if we compare additionally to this special field and both are true, we can trust it. What do you think?

frikky · 2026-03-12T21:34:47Z


-		returnAction, err := HandleAiAgentExecutionStart(workflowExecution, actionResult.Action, true)
+		var originalAction Action
+		if foundActionResultIndex >= 0 && foundActionResultIndex < len(workflowExecution.Results) {


What was the point of this? I'm always confused when we override actions, when things used to work without it

At line 8295 in ai.go, we are building actionResult, and inside of that we are setting the action's label name as fmt.Sprintf("Agent Decision %s", decision.RunDetails.Id). Then we call handleAgentDecisionStreamResult, and that one passes actionResult to HandleAiAgentExecutionStart. Now startNode is overwritten with an action with label, e.g., "Agent Decision RNnBU2aY". Due to this, the later nodes in workflow that are trying to refer to the AI agent node with its original label can't find it, resulting in being unable to get any data from the agent node at all.

satti-hari-krishna-reddy · 2026-04-24T09:26:18Z

@copilot explain the changes made in this PR

Copilot

Pull request overview

This PR tightens up AI agent execution lifecycle handling by adding trace/caller metadata propagation, improving abort/failure state reporting, and reducing duplicate/self-request/log noise via caching.

Changes:

Add X-Internal-Caller / X-Trace-ID support (CORS + internal requests) and thread context.Context through agent/automation entry points.
Adjust agent self-request caching TTLs and reduce repeated “abort due to result failure” logging via cache-based de-duplication.
Refactor agent early termination into abortAgentExecution, improve LLM error logging, and add recovery behavior when LLM returns malformed decision JSON.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 5 comments.

File	Description
shared.go	Adds CORS headers, tweaks agent self-request caching TTL, propagates `ctx` into agent start paths.
db-connector.go	Adds cache-based log de-dupe for abort logging; passes `ctx` into datastore automation runner.
codegen.go	Threads `ctx` into datastore automation runner; adds trace-id propagation headers to internal agent start request.
ai.go	Introduces `abortAgentExecution`, changes abort semantics to `ABORTED`, improves LLM failure handling/logging and decision parsing recovery.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-24T17:11:10Z

+    	traceID, _ := ctx.Value("trace_id").(string)
+		if strings.TrimSpace(caller) == "" {
+			log.Printf("ERROR[%s] AI agent: No caller function info provided for AI agent, aborting the request ...", execution.ExecutionId)
+			return abortAgentExecution(ctx, execution, startNode, AgentOutput{}, "no_caller_info", "No caller function info provided for AI Agent start")


HandleAiAgentExecutionStart now aborts if ctx is missing a non-empty "caller" value. In this repo, GetContext(request) currently always returns context.Background() (no values), so this change will cause agent starts to abort in normal request flows unless every call site/middleware injects ctx.Value("caller"). Consider making caller optional (log a warning + default), or ensure all entry points populate it (e.g., derive from an HTTP header like X-Internal-Caller, or set it at the immediate call sites).

Suggested change

traceID, _ := ctx.Value("trace_id").(string)

if strings.TrimSpace(caller) == "" {

log.Printf("ERROR[%s] AI agent: No caller function info provided for AI agent, aborting the request ...", execution.ExecutionId)

return abortAgentExecution(ctx, execution, startNode, AgentOutput{}, "no_caller_info", "No caller function info provided for AI Agent start")

traceID, _ := ctx.Value("trace_id").(string)

if strings.TrimSpace(caller) == "" {

caller = "unknown"

log.Printf("[WARNING][%s] AI agent: No caller function info provided for AI agent start; defaulting caller=%q", execution.ExecutionId, caller)

Copilot · 2026-04-24T17:11:11Z

+				b := make([]byte, 6)
+				rand.Read(b)
+
+				mappedDecisions = []AgentDecision{{
+					I:                lastFinishedIndex,
+					Category:         "finish",
+					Action:           "finish",
+					Tool:             "core",
+					Confidence:       1.0,
+					Runs:             "1",
+					ApprovalRequired: false,
+					Reason:           reason,
+					Fields: []Valuereplace{{
+						Key:   "output",
+						Value: outputValue,
+					}},
+					RunDetails: AgentDecisionRunDetails{
+						Id:          base64.RawURLEncoding.EncodeToString(b),
+						Status:      "",


rand.Read here is from math/rand (not crypto/rand) and the global RNG is never seeded in this file, so the generated RunDetails.Id can be deterministic across process starts and collide. It also ignores the returned error/byte count. Prefer using crypto/rand (with error handling) or an existing UUID generator (uuid.NewV4) to create unique, non-deterministic IDs for decision/run identifiers.

Copilot · 2026-04-24T17:11:11Z

+	if ctx == nil {
+        ctx = context.Background() 
+    }


The ctx==nil guard is a good idea, but the indentation/trailing whitespace here looks inconsistent with gofmt formatting. Running gofmt on this block will keep formatting consistent across the file.

Co-authored-by: Copilot <copilot@github.com>

satti-hari-krishna-reddy requested a review from frikky March 12, 2026 14:25

frikky reviewed Mar 12, 2026

View reviewed changes

satti-hari-krishna-reddy requested a review from frikky March 18, 2026 17:20

satti-hari-krishna-reddy marked this pull request as draft April 1, 2026 08:29

satti-hari-krishna-reddy force-pushed the timeout-fix branch from eddf2c8 to cff8a26 Compare April 1, 2026 16:53

satti-hari-krishna-reddy marked this pull request as ready for review April 6, 2026 06:29

satti-hari-krishna-reddy changed the title ~~Fix agent stuck in waiting state after completion~~ Lot's of agent stuff again Apr 6, 2026

satti-hari-krishna-reddy mentioned this pull request Apr 6, 2026

fix: handle raw LLM content with fallback #351

Draft

satti-hari-krishna-reddy self-assigned this Apr 10, 2026

satti-hari-krishna-reddy requested a review from Copilot April 24, 2026 17:05

Copilot started reviewing on behalf of satti-hari-krishna-reddy April 24, 2026 17:05 View session

Copilot AI reviewed Apr 24, 2026

View reviewed changes

satti-hari-krishna-reddy added 17 commits April 27, 2026 10:54

add http app

dca3acb

fix agent stuck loading due to timeout

55e6a27

added auth injection for other internal apps as well

1d7c968

include more conditions

db9f83b

commented out not used code

8a2855c

a ton of changes again

3acdb25

undo some changes for now

04deb5b

increase the TTL terminal statuses

44924c5

fixed some abort bugs

aa044c1

general output format fix

44c8398

added more visibility in logs

26de432

added trace logs to handleRunDatastoreAutomation

9148a37

using the context instead of modifying function signature

943e2e7

use the context for passing caller and trace info

117b3f5

update the call with right params

1075052

commented out for now

e50fef7

added new alowed headers to cors check

3275f96

code ql fix: added Escape single quotes

1e556f8

satti-hari-krishna-reddy force-pushed the timeout-fix branch from 0c4df21 to 1e556f8 Compare April 27, 2026 05:32

satti-hari-krishna-reddy and others added 2 commits April 27, 2026 14:06

fix: ensure caller context is set in PrepareSingleAction

d60c6f9

Co-authored-by: Copilot <copilot@github.com>

fix: replace random trace ID generation with UUID

10cc3c3

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Lot's of agent stuff again#343

Lot's of agent stuff again#343
satti-hari-krishna-reddy wants to merge 20 commits intoShuffle:mainfrom
satti-hari-krishna-reddy:timeout-fix

satti-hari-krishna-reddy commented Mar 12, 2026 •

edited

Loading

Uh oh!

frikky Mar 12, 2026

Uh oh!

satti-hari-krishna-reddy Mar 18, 2026

Uh oh!

frikky Mar 12, 2026

Uh oh!

satti-hari-krishna-reddy Mar 18, 2026 •

edited

Loading

Uh oh!

satti-hari-krishna-reddy commented Apr 24, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 24, 2026

Uh oh!

Uh oh!

Copilot AI Apr 24, 2026

Uh oh!

Uh oh!

Copilot AI Apr 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

satti-hari-krishna-reddy commented Mar 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

frikky Mar 12, 2026

Choose a reason for hiding this comment

Uh oh!

satti-hari-krishna-reddy Mar 18, 2026

Choose a reason for hiding this comment

Uh oh!

frikky Mar 12, 2026

Choose a reason for hiding this comment

Uh oh!

satti-hari-krishna-reddy Mar 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

satti-hari-krishna-reddy commented Apr 24, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 24, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI Apr 24, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI Apr 24, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

satti-hari-krishna-reddy commented Mar 12, 2026 •

edited

Loading

satti-hari-krishna-reddy Mar 18, 2026 •

edited

Loading