SSE implementation #767

Sandeep-BA · 2026-01-23T18:37:15Z

No description provided.

feat: Introduced ApplicationIntegrationToolset in JavaADK

Merge

…g both streaming and non-streaming capabilities

feat: Add BaseToolset and update McpToolset to use the new interface

Merge

merge

Integration of Amazon Bedrock for Claude Models. Example and readme.

Merge

google adk merge

Long Term Session memory

adk-core google

… BaseMemoryService constructor

Reverting CassandraRagRetrival extra chnages

Google adk merge

…, Remove DB credential constants, use literals directly, Reduce ConnectionTimeout, HikariCP connection for high-performance connection pooling

Updated the PostgresArtifactService

Google fork syncing

Removed create table for artifacts

…ative) - Make HttpServer SSE default endpoint on port 9085 - Add Spring SSE alternative endpoint on port 9086 - Fix JSON parsing: Change from Gson to Jackson ObjectMapper - Add comprehensive framework comparison documentation - Add testing scripts and documentation - Add unit and integration tests Changes: - HttpServerSseController: Use Jackson ObjectMapper for JSON parsing - HttpServerSseConfig: Default port 9085, enabled by default - ExecutionController: Spring endpoint renamed to /run_sse_spring - application.properties: Configure Spring port 9086, HttpServer port 9085 Documentation: - SSE_FRAMEWORK_COMPARISON.md: Comprehensive framework comparison - TEST_SSE_ENDPOINT.md: Testing guide - QUICK_START_SSE.md: Quick start guide - COMMIT_GUIDE.md: Commit instructions - TEST_RESULTS.md: Test results - TESTING_SUMMARY.md: Test summary Tests: - HttpServerSseControllerTest: Unit tests - HttpServerSseControllerIntegrationTest: Integration tests - Updated existing SseEventStreamService tests Author: Sandeep Belgavi Date: January 24, 2026

gemini-code-assist · 2026-01-23T18:39:07Z

Summary of Changes

Hello @Sandeep-BA, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly expands the Agent Development Kit (ADK) for Java by introducing a flexible and performant Server-Sent Events (SSE) streaming mechanism, offering both a lightweight Java HttpServer option and a feature-rich Spring-based alternative. Concurrently, it integrates a diverse set of persistence solutions including PostgreSQL, MapDB, MongoDB, Redis, and Cassandra for robust storage of artifacts and session data. The changes also broaden LLM model compatibility with new Bedrock and Ollama integrations, enhance Gemma's function calling, and introduce a Kafka-based event processing pipeline, thereby establishing a comprehensive multi-tier architecture for scalable and resilient agent development.

Highlights

Server-Sent Events (SSE) Implementation: Introduced two distinct SSE implementations: a lightweight Java HttpServer-based option (default on port 8081) and a Spring-based alternative (on port 8080). Both can run simultaneously and share the same request/response format.
Enhanced Persistence Layer: Integrated multiple persistence solutions including PostgreSQL, MapDB, MongoDB, Redis, and Cassandra for storing artifacts and session data, offering diverse choices for different use cases.
Expanded LLM Model Support: Added new integrations for Bedrock and Ollama LLMs, enhanced Gemma with function calling emulation, and introduced Redbus AD Gateway for Azure LLM access, broadening the range of supported models.
Kafka Event Processing Pipeline: Implemented a Kafka consumer service to process ADK events in real-time and persist them to PostgreSQL, supporting a robust multi-tier architecture for event streaming and data consistency.
Comprehensive Documentation: Provided extensive new Markdown documentation covering SSE implementation details, framework comparisons, quick start guides, testing procedures, and architectural overviews for the new features.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

…ative) - Make HttpServer SSE default endpoint on port 9085 - Add Spring SSE alternative endpoint on port 9086 - Fix JSON parsing: Change from Gson to Jackson ObjectMapper - Add comprehensive guide with pros/cons and usage instructions - Add testing scripts and unit/integration tests Changes: - HttpServerSseController: Use Jackson ObjectMapper for JSON parsing - HttpServerSseConfig: Default port 9085, enabled by default - ExecutionController: Spring endpoint renamed to /run_sse_spring - application.properties: Configure Spring port 9086, HttpServer port 9085 Documentation: - SSE_GUIDE.md: Comprehensive guide with pros/cons and usage Tests: - HttpServerSseControllerTest: Unit tests - HttpServerSseControllerIntegrationTest: Integration tests - Updated existing SseEventStreamService tests Author: Sandeep Belgavi Date: January 24, 2026

gemini-code-assist

Code Review

This pull request implements SSE with HttpServer (default) and Spring (alternative), including implementation, configuration, testing, and documentation. The code review highlights several issues, including potential data loss due to incorrect primary key usage in the event_content_parts table, fragile string splitting for filenames and versions, non-atomic operations in Redis, blocking HTTP requests, and the need for more robust error handling. It also suggests improvements to configuration management, JSON handling, and adherence to Bedrock API specifications.

gemini-code-assist · 2026-01-23T19:03:09Z

README.md

+LlmAgent.builder()
+            .name(NAME)
+            .model(new com.google.adk.models.OllamaBaseLM("qwen3:0.6b"))//.model(new RedbusADG("40"))
+            .description("Agent to calculate trigonometric functions (sine, cosine, tangent) for given angles.") // Updated description


Using event_id as the PRIMARY KEY for the event_content_parts table implies that each event can only have one content part. If an event can have multiple parts (e.g., text, image, function call), then event_id alone cannot uniquely identify each row. This would lead to data loss, as only the last part inserted for a given event_id would be retained due to the ON CONFLICT (event_id) DO UPDATE SET clause in KafkaEventRepository. Consider using a composite primary key, such as (event_id, part_index) or (event_id, part_type) if part_type is unique per event, or introduce a separate unique part_id column.

gemini-code-assist · 2026-01-23T19:03:09Z

core/src/main/java/com/google/adk/kafka/repository/KafkaEventRepository.java

+
+      for (KafkaEventData.Part part : parts) {
+        String partType = part.getPartType();
+        String textContent = null;
+        String functionCallId = null;
+        String functionCallName = null;
+        String functionCallArgs = null;
+        String functionResponseId = null;
+        String functionResponseName = null;
+        String functionResponseData = null;
+
+        switch (partType) {
+          case "text":
+            textContent = part.getText();
+            break;
+          case "functionCall":
+            if (part.getFunctionCall() != null) {
+              functionCallId = part.getFunctionCall().getId();
+              functionCallName = part.getFunctionCall().getName();
+              functionCallArgs = part.getFunctionCall().getArgs();
+            }
+            break;
+          case "functionResponse":
+            if (part.getFunctionResponse() != null) {
+              functionResponseId = part.getFunctionResponse().getId();
+              functionResponseName = part.getFunctionResponse().getName();
+              functionResponseData = part.getFunctionResponse().getResponse();
+            }
+            break;
+          default:
+            logger.warn("Unknown content part type for event ID: {} - {}", eventId, partType);
+            continue;
+        }
+
+        stmt.setString(1, eventId);
+        stmt.setString(2, sessionId);
+        stmt.setString(3, partType);
+        stmt.setString(4, textContent);
+        stmt.setString(5, functionCallId);
+        stmt.setString(6, functionCallName);
+        stmt.setString(7, safeJsonString(functionCallArgs));
+        stmt.setString(8, functionResponseId);
+        stmt.setString(9, functionResponseName);
+        stmt.setString(10, safeJsonString(functionResponseData));
+
+        stmt.addBatch();
+      }
+
+      int[] rowsAffected = stmt.executeBatch();
+      conn.commit();


The event_content_parts table schema defines event_id as the PRIMARY KEY. However, the upsertEventContentParts method iterates through a List<KafkaEventData.Part> and attempts to insert each part using the same event_id. Due to the ON CONFLICT (event_id) DO UPDATE SET clause, this will result in only the last part from the list being stored for a given event_id, effectively discarding all other parts. If an event can have multiple content parts, the primary key for event_content_parts must be a composite key (e.g., (event_id, part_index) or (event_id, part_type) if part_type is unique per event) to allow multiple entries per event.

gemini-code-assist · 2026-01-23T19:03:09Z

core/src/main/java/com/google/adk/models/BedrockBaseLM.java

+                    .required()
+                    .ifPresent(requiredList -> parametersMap.put("required", requiredList));
+                toolMap.put("parameters", parametersMap);
+              }


The tool schema is being put under the key "parameters" in the toolMap. However, the Bedrock converse API expects the tool schema to be directly under "inputSchema" within the toolSpec object, not nested under "parameters". This incorrect structure will likely lead to API errors when attempting to use tool calling. Please adjust the JSON structure to match Bedrock's expected format for tool declarations.

gemini-code-assist · 2026-01-23T19:03:09Z

core/pom.xml

+      <groupId>org.testcontainers</groupId>
+      <artifactId>cassandra</artifactId>
+      <version>1.19.7</version>
+      <scope>test</scope>
+    </dependency>
+    <dependency>
+      <groupId>org.testcontainers</groupId>
+      <artifactId>junit-jupiter</artifactId>
+      <version>1.19.7</version>
+      <scope>test</scope>
+    </dependency>


The testcontainers dependencies (cassandra and junit-jupiter) are correctly scoped as test. However, the testcontainers dependency itself is not explicitly declared with test scope, which means it will be included in the runtime classpath of the core module. Test-scoped dependencies should not be part of the main application's runtime dependencies.

gemini-code-assist · 2026-01-23T19:03:09Z

core/src/main/java/com/google/adk/models/RedbusADG.java

+          httpClient.send(httpRequest, HttpResponse.BodyHandlers.ofInputStream());
+      int statusCode = response.statusCode();


The httpClient.send() method is a blocking operation. In a streaming context, blocking the thread while waiting for the entire response to be received can negate the benefits of streaming and lead to performance bottlenecks. For true streaming, HttpClient.sendAsync() should be used, and the HttpResponse.BodyHandlers.ofInputStream() should be processed asynchronously as data arrives.

gemini-code-assist · 2026-01-23T19:03:11Z

core/src/main/java/com/google/adk/utils/PostgresDBHelper.java

+      }
+
+      JSONObject corrected = new JSONObject("{" + "\"events\":" + raw + "}");
+      JSONArray events = corrected.getJSONArray("events");


The line `JSONObject corrected = new JSONObject("{" + ""events":

gemini-code-assist · 2026-01-23T19:03:11Z

core/src/main/java/com/google/adk/utils/PostgresDBHelper.java

+                      + "session_id=EXCLUDED.session_id, author=EXCLUDED.author, actions_state_delta=EXCLUDED.actions_state_delta, "
+                      + "actions_artifact_delta=EXCLUDED.actions_artifact_delta, actions_requested_auth_configs=EXCLUDED.actions_requested_auth_configs, "
+                      + "actions_transfer_to_agent=EXCLUDED.actions_transfer_to_agent, content_role=EXCLUDED.content_role, "
+                      + "timestamp=EXCLUDED.timestamp, invocation_id=EXCLUDED.invocation_id");


The line stmtEvents.setString(4, actions.opt("stateDelta").toString()); uses opt("stateDelta").toString() to get the JSON string for a JSONB column. If stateDelta is a complex JSON object, toString() might not produce a valid JSON string suitable for direct insertion into a JSONB column, or it might produce a string with unwanted escaping. It's safer to explicitly serialize the stateDelta object to a JSON string using objectMapper.writeValueAsString(actions.optJSONObject("stateDelta")) to ensure correct JSON formatting.

gemini-code-assist · 2026-01-23T19:03:11Z

core/src/main/java/com/google/adk/utils/PostgresDBHelper.java

+        for (int i = 0; i < events.length(); i++) {
+          JSONObject ev = events.getJSONObject(i);
+          //          logger.info("Processing event for insertion: {}", ev.toString());
+          JSONObject actions = ev.getJSONObject("actions");


The line stmtParts.setString(7, fc != null ? fc.opt("args").toString() : null); uses fc.opt("args").toString() to get the JSON string for a JSONB column. If args is a complex JSON object, toString() might not produce a valid JSON string suitable for direct insertion into a JSONB column, or it might produce a string with unwanted escaping. It's safer to explicitly serialize the args object to a JSON string using objectMapper.writeValueAsString(fc.optJSONObject("args")) to ensure correct JSON formatting.

gemini-code-assist · 2026-01-23T19:03:11Z

core/src/main/java/com/google/adk/utils/PostgresDBHelper.java

+            stmtParts.executeUpdate();
+          }
+        }
+        conn.commit();


The conn.commit() is called inside the loop, meaning a separate transaction is committed for each event. This is inefficient if there are many events to insert for a single session, as each commit incurs overhead. It would be more efficient to commit once after all events for a session have been inserted, or to use batch updates and commit the entire batch.

gemini-code-assist · 2026-01-23T19:03:11Z

README.md

+            log.info("Using MapDbSessionService");
+            return new MapDbSessionService("map.db");
+        } catch (Exception ex) {
+            java.util.logging.Logger.getLogger(AdkWebServer.class.getName()).log(Level.SEVERE, null, ex);


The default value for event_data is an empty JSON array []. However, the KafkaEventRepository's parseEventData method expects event_data to contain an events array within a JSON object (e.g., {"events": [...]}). This discrepancy might lead to parsing errors or unexpected behavior if event_data is initialized with this default and later processed. Consider if event_data should default to {}, or if the parsing logic should explicitly handle an empty array as a valid input for an empty list of events.

manojkumarredbus and others added 30 commits July 7, 2025 12:42

Merge pull request #11 from google/main

3b19510

feat: Introduced ApplicationIntegrationToolset in JavaADK

Merge branch 'main' of https://github.com/redbus-labs/adk-java

00d0aff

started to add streaming support

e2fed4e

Merge branch 'main' of https://github.com/redbus-labs/adk-java

e1bc622

Merge pull request #12 from google/main

5e4e772

Merge

Streaming works

3025285

Streaming works

274b18a

Merge branch 'main' of https://github.com/redbus-labs/adk-java

0320203

Update README.md

4c4135b

Function args parsing fixed for streaming

e99bc83

Merge branch 'main' of https://github.com/redbus-labs/adk-java

203f565

Merge pull request #13 from google/main

6e9c55a

Merge

The changes provide a more robust and compatible behavior for handlin…

696822f

…g both streaming and non-streaming capabilities

Merge branch 'main' of https://github.com/redbus-labs/adk-java

29101ca

Merge pull request #14 from google/main

bd9903c

feat: Add BaseToolset and update McpToolset to use the new interface

changes related to audio streamin

5d3946b

RedbusADG streaming implementation.

55a46b4

Merge branch 'main' of https://github.com/redbus-labs/adk-java

44541fb

Merge pull request #15 from google/main

d0a955b

Merge

RedbusADG streaming implementation. Fixing streaming

8cd29ae

Merge branch 'main' of https://github.com/redbus-labs/adk-java

0e351f0

merge conflict resolve

135cc97

RedbusADG streaming implementation. Fixing streaming

de9494e

resolve merge conflict

eb8cb58

Update README.md

ebf7689

Merge pull request #16 from google/main

eb9a8d0

merge

Merge branch 'main' of https://github.com/redbus-labs/adk-java

ec0c2fd

Integration of Amazon Bedrock for Claude Models. Example and readme.

e3c438a

Merge pull request #17 from akshayar/main

f4df9fc

Integration of Amazon Bedrock for Claude Models. Example and readme.

Merge pull request #18 from google/main

8effc9f

Merge

Yashas Shetty and others added 25 commits January 5, 2026 16:10

removed readme file and reverted change in PostgresSessionService

2210518

Merge pull request #112 from google/main

95357c3

google adk merge

fixed log and limit

e52471b

Merge pull request #111 from pawanshaiitd/feature/adk-memory

928f04d

Long Term Session memory

changes related to gemma , that eventually don't work due to 400 error

4c3d763

Merge branch 'main' of https://github.com/redbus-labs/adk-java

4a42554

Ollama support for image models

6aebd67

Update README.md

d3378e5

addressed the review comments and made the required changes

8b1ce1c

Merge pull request #113 from google/main

c03542e

adk-core google

Resolve merge conflict: keep PostgresArtifactService spelling + merge…

92fb917

… BaseMemoryService constructor

Merge branch 'main' into main

9937ad7

Reverting CassandraRagRetrival extra chnages

b29cb3d

Merge pull request #115 from pawanshaiitd/fix/MemoryTestCases

9e3fbf1

Reverting CassandraRagRetrival extra chnages

Merge pull request #114 from google/main

e7a9d85

Google adk merge

Clean up environment variable handling and adjust connection timeout…

95e1ce3

…, Remove DB credential constants, use literals directly, Reduce ConnectionTimeout, HikariCP connection for high-performance connection pooling

Merge pull request #108 from yashas2604/main

c22dd66

Updated the PostgresArtifactService

Merge upstream main branch

a7786c6

Fixed the failing UTs

206ace7

Merge pull request #117 from google/main

b7e805e

Google fork syncing

removed create table for artifacts

af8ccc6

Merge pull request #120 from yashas2604/main

7c9e833

Removed create table for artifacts

Merging upstream main branch from google adk

0bf5eec

Fixing build failure

3d1a2ce

Sandeep-BA force-pushed the sse branch from 4c6c2f4 to c0c41a9 Compare January 23, 2026 18:39

Sandeep-BA closed this Jan 23, 2026

gemini-code-assist bot reviewed Jan 23, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

SSE implementation #767

SSE implementation #767

Sandeep-BA commented Jan 23, 2026

Uh oh!

gemini-code-assist bot commented Jan 23, 2026

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

gemini-code-assist bot Jan 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

		httpClient.send(httpRequest, HttpResponse.BodyHandlers.ofInputStream());
		int statusCode = response.statusCode();

SSE implementation #767

SSE implementation #767

Conversation

Sandeep-BA commented Jan 23, 2026

Uh oh!

gemini-code-assist bot commented Jan 23, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 23, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants