feat: Add IBM Watsonx and Ollama to Agent Component

[edwinjosechittilappilly]

This pull request updates the Instagram Copywriter.json starter project to expand support for additional AI providers and improve configuration options. The most significant changes include the addition of new provider options and a new API base URL setting.

Provider support enhancements:

• Added IBM watsonx.ai and Ollama to the list of available provider options, allowing users to select these new AI services in addition to existing ones.

Configuration improvements:

• Introduced a new base_url field to the configuration, enabling users to specify the API base URL for their selected provider. This makes the setup more flexible and adaptable to different environments.

Metadata update:

• Updated the code_hash value in the metadata to reflect the new changes in the configuration.

Summary by CodeRabbit

• New Features
  • Added IBM watsonx.ai and Ollama as new model provider options for agent configuration across all starter projects.
  • Introduced new configurable fields for agents: API base URL, project ID, and maximum output token limits.
  • Enhanced model provider selection and agent setup configuration UI across starter projects.

[coderabbitai]

Important

Review skipped

Auto incremental reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Walkthrough

Starter projects are updated with three new Agent inputs (base_url, project_id, max_output_tokens) and expanded model provider options for IBM watsonx.ai and Ollama. Knowledge base components refactored for lazy-loaded root path resolution. Model provider constants extended with new providers and conditional input processing. Agent, Watsonx, and Ollama components updated with new/renamed public inputs.

Changes

Cohort / File(s)	Summary
Starter Projects - Agent Configuration src/backend/base/langflow/initial_setup/starter_projects/Instagram Copywriter.json, Invoice Summarizer.json, Market Research.json, News Aggregator.json, Nvidia Remix.json, Pokédex Agent.json, Price Deal Finder.json, Research Agent.json, SaaS Pricing.json, Search agent.json, Simple Agent.json, Social Media Agent.json, Youtube Analysis.json	Added three public inputs to Agent components: base_url (StrInput), project_id (StrInput), max_output_tokens (IntInput). Expanded model provider dropdown options to include IBM watsonx.ai and Ollama. Updated Agent component code_hash metadata from bdb8b8db6375 to 6f2517926f3d across projects.
Knowledge Base Path Resolution Refactoring src/backend/base/langflow/initial_setup/starter_projects/Knowledge Ingestion.json, Knowledge Retrieval.json	Refactored knowledge base root path handling with lazy-loaded initialization via _get_knowledge_bases_root_path() helper and global _KNOWLEDGE_BASES_ROOT_PATH cache. Replaced direct root path references with centralized helper function for improved path resolution and error handling.
Model Provider Support src/lfx/src/lfx/base/models/model_input_constants.py	Extended MODEL_PROVIDERS_DICT with IBM watsonx.ai and Ollama entries including fields, inputs, component classes, and icons. Added provider-specific input processing via _get_ollama_inputs_and_fields() and _get_watsonx_inputs_and_fields() helpers. Updated get_filtered_inputs() and process_inputs() to accept optional provider_name parameter for conditional input modification.
Agent Component Input Expansion src/lfx/src/lfx/components/agents/agent.py	Added new public inputs: api_key (SecretStrInput), base_url (StrInput), project_id (StrInput), max_output_tokens (IntInput). Strengthened delete_fields() method with None-safety checks for build_config.
Watsonx Component Refactoring src/lfx/src/lfx/components/ibm/watsonx.py	Renamed dropdown input from url to base_url with hard-coded list of API endpoints. Updated Model Name input to use real_time_refresh with refresh button. Modified update_build_config() to react to base_url and model_name field changes with dictionary-based access. Updated build_model() to use self.base_url instead of self.url.
Ollama Component Input Renaming src/lfx/src/lfx/components/ollama/ollama.py	Renamed BoolInput from verbose to enable_verbose_output with display name "Ollama Verbose Output". Updated llm_params mapping to use renamed field in build_model().

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

The PR spans multiple heterogeneous changes across diverse file types: 13 starter project JSON files with similar but distinct input/provider updates requiring individual validation, two knowledge base components with refactored path resolution logic, model provider constants with new conditional logic, and component-level input/parameter modifications. While many changes follow repetitive patterns (particularly in starter projects), each requires validation for correctness, and the provider-specific logic in model_input_constants and component refactoring demand careful review.

Possibly related PRs

• fix(setup): preserve advanced attribute and generalize starter project JSON #8529 — Starter project setup changes that skip updating the "advanced" attribute align with fields marked as "advanced" in the new inputs.
• refactor: Update AgentComponent to utilize MODEL_OPTIONS_METADATA from constants #9969 — Modifies AgentComponent provider metadata and model-provider imports using shared MODEL_* constants, similar to changes in this PR.
• fix: disable mcp auto install for not installed applications, refactor mcp-projects logic #9599 — Modifies knowledge-bases ingestion/retrieval components with path resolution and refactoring patterns similar to the Knowledge Ingestion/Retrieval JSON changes.

Suggested labels

enhancement, size:M, provider-support

Suggested reviewers

• lucaseduoli

Pre-merge checks and finishing touches

❌ Failed checks (1 error, 4 warnings)

Check name	Status	Explanation	Resolution
Test Coverage For New Implementations	❌ Error	Based on my analysis of the PR, the pull request adds significant new functionality including support for IBM watsonx.ai and Ollama providers, new input fields (base_url, project_id, max_output_tokens) to the Agent component, and modifications to multiple starter project JSON files. However, the PR does not include corresponding test files for these changes. The modifications include: (1) new helper functions in model_input_constants.py for handling provider-specific inputs, (2) new input declarations in the Agent component, (3) changes to watsonx.py and ollama.py components, and (4) updates to 15 starter project JSON files. While the raw summary identifies multiple locations where tests should be added (particularly for provider-specific field visibility, conditional showing/hiding based on selected provider, and new input handling), no new test files appear to have been added to address these changes, nor are there existing comprehensive tests for the Agent component's update_build_config behavior or model input filtering logic.	The PR should include test files for: (1) src/lfx/src/lfx/base/models/model_input_constants.py - tests for _get_ollama_inputs_and_fields() and _get_watsonx_inputs_and_fields() functions, and tests for process_inputs() with provider_name parameter; (2) src/lfx/src/lfx/components/agents/agent.py - tests for the agent's update_build_config method to verify conditional visibility of base_url, project_id, and max_output_tokens based on selected provider; (3) src/lfx/src/lfx/components/ibm/watsonx.py - tests for the updated dropdown input handling and base_url parameter changes; (4) src/lfx/src/lfx/components/ollama/ollama.py - tests for the renamed enable_verbose_output field; and (5) integration tests to verify the starter projects load correctly without validation errors when using default providers. Additionally, regression tests should be added to ensure existing OpenAI provider functionality is not broken by these changes.
Docstring Coverage	⚠️ Warning	Docstring coverage is 36.36% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.
Test Quality And Coverage	⚠️ Warning	The PR (branch coderabbit_10357, commit c805135 "Update Ollama model input constants and logic") introduces substantial new functionality including IBM watsonx.ai and Ollama provider support across model_input_constants.py, agent.py, watsonx.py, ollama.py, and 14 starter project JSON files, with updates to function signatures and new public input fields. However, the git log confirms "No test files in recent commit" for this PR. While the repository has well-established testing patterns with comprehensive existing tests (test_agent_component.py with 615 lines, test_chatollama_component.py with extensive async mocking and proper pytest patterns), the absence of new tests in this PR means none of the new provider integrations, conditional field visibility logic, async method changes, or the critical bugs identified in review comments (debug print statements, type mismatches, required-but-hidden fields) are validated by tests before merge.	The PR must include comprehensive test coverage before merge following the project's established patterns: (1) add unit tests in test_model_input_constants.py for get_filtered_inputs and process_inputs with provider_name parameter, testing all provider variants (OpenAI, IBM watsonx.ai, Ollama, None) and ImportError edge cases; (2) extend test_agent_component.py with async test methods using @pytest.mark.asyncio validating conditional field visibility (base_url, project_id, max_output_tokens shown/hidden based on provider selection) across all supported providers; (3) add tests for watsonx.py covering the url-to-base_url refactoring, dropdown initialization, and model refresh logic using AsyncMock patterns like those in test_chatollama_component.py; (4) verify ollama.py changes for the enable_verbose_output field mapping; and (5) add integration tests confirming that starter projects load without validation errors and that all reviewed bugs (debug prints removed, type mismatches fixed, conditional field visibility working) are resolved before merge.
Test File Naming And Structure	⚠️ Warning	The PR makes substantial changes to multiple core Python components (agent.py, watsonx.py, ollama.py, model_input_constants.py) and multiple JSON starter project files, but the raw_summary indicates no test files were added or modified. The repository has an established testing pattern with 268 existing test files using pytest conventions. Specifically, existing tests exist for Agent components (test_agent_component.py, 615 lines) and Ollama components (test_chatollama_component.py, 459 lines), but searching for the new provider-specific fields (watsonx, ollama, base_url, project_id, max_output_tokens) in the existing agent test yields zero results, indicating these critical new features have no test coverage. No test file exists for the Watsonx component despite it being substantially modified, and no tests exist for the model_input_constants.py changes that add new provider entries and modify function signatures to include a provider_name parameter. The existing test patterns in the codebase use descriptive async test methods, proper fixtures, and mocking, which should be replicated for the new functionality.	The PR must add test files following the established pytest patterns in the repository. Specifically: (1) create ./src/backend/tests/unit/components/ibm/test_watsonx_component.py with descriptive test functions (test_build_model, test_update_build_config_base_url, test_fetch_models_success, test_fetch_models_failure) following the Ollama test structure; (2) add test cases to the existing test_agent_component.py with descriptive names like test_provider_specific_base_url_visibility, test_provider_specific_project_id_visibility, and test_max_output_tokens_field_initialization to verify the new fields and conditional visibility based on selected provider; (3) add ./src/backend/tests/unit/base/models/test_model_input_constants.py testing get_filtered_inputs and process_inputs with provider_name parameter, new provider entries for Watsonx and Ollama, and MODEL_PROVIDERS_DICT updates; (4) update existing test_chatollama_component.py to verify the renamed enable_verbose_output field. All tests should use async test methods, pytest fixtures, mocking where appropriate, and cover both positive scenarios (fields present, correct values) and negative scenarios (missing providers, API failures, invalid configurations).
Excessive Mock Usage Warning	⚠️ Warning	Analysis of the existing test suite reveals that test files for related components follow appropriate mocking practices. The test_chatollama_component.py file contains 26 @patch decorators and 43 mock initializations across 18 test methods (2.39 mocks per test), but these mocks are appropriately targeted at external dependencies (ChatOllama, httpx.AsyncClient, socket, filesystem operations) rather than internal logic, representing best practices. In contrast, test_agent_component.py uses minimal mocking (0.04 ratio) and primarily tests real component behavior. However, this PR introduces significant new functionality (IBM Watsonx integration, new input fields base_url/project_id/max_output_tokens, provider-specific conditional logic) without adding any corresponding test files or modifications to existing tests. The absence of test coverage for these changes means the custom check for excessive mock usage cannot assess the new code—but more importantly, the lack of any test additions suggests inadequate testing practices that could lead to poor test quality if tests are later added without proper guidance.	This PR fails the excessive mock usage check by omission: it adds new provider integrations and component logic without corresponding test coverage. To pass this check, add unit tests following the patterns demonstrated in test_chatollama_component.py—mock only external dependencies (API calls to Watsonx, network requests) while testing real component logic and state transitions. Create specific tests for: (1) provider-specific field visibility in update_build_config, (2) watsonx component initialization and model selection, (3) conditional required/show flags based on selected provider, and (4) integration tests validating end-to-end provider switching. Ensure mock-to-test ratios remain reasonable (under 3 mocks per test) by grouping related mocks and testing actual business logic rather than mocking core functionality.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title Check	✅ Passed	The pull request title "feat: Add IBM Watsonx and Ollama to Agent Component" directly reflects a significant and concrete aspect of the changeset. The changes do indeed add IBM watsonx.ai and Ollama as new model provider options across multiple starter projects and backend components. However, the title is somewhat narrow in scope—while it accurately captures the provider expansion, it does not mention the equally significant configuration enhancements (base_url, project_id, max_output_tokens fields) that are introduced alongside the provider support, nor does it reflect that these changes apply broadly across 15+ starter projects and backend infrastructure beyond just the Agent Component.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 24

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (5)

src/backend/base/langflow/initial_setup/starter_projects/Instagram Copywriter.json (1)

2090-2107: Add metadata entries for new provider options.

The Agent model provider dropdown now includes "IBM watsonx.ai" and "Ollama" (lines 2090-2095), but the options_metadata array (lines 2097-2107) only has three entries corresponding to "Anthropic", "Google Generative AI", and "OpenAI". This mismatch may cause UI rendering issues or missing provider icons when the new options are selected.

Extend the options_metadata array to include metadata objects for "IBM watsonx.ai" and "Ollama":

                "options_metadata": [
                  {
                    "icon": "Anthropic"
                  },
                  {
                    "icon": "GoogleGenerativeAI"
                  },
                  {
                    "icon": "OpenAI"
+                 },
+                 {
+                   "icon": "IBMWatsonx"
+                 },
+                 {
+                   "icon": "Ollama"
                  }
                ],

Verify the exact icon names match the icon registry; use "IBMWatsonx" and "Ollama" as placeholders if you need to confirm the correct identifiers.

src/backend/base/langflow/initial_setup/starter_projects/Market Research.json (1)

1658-1678: Provider options and metadata count mismatch.

The agent_llm dropdown defines 5 provider options (Anthropic, Google Generative AI, OpenAI, IBM watsonx.ai, Ollama) but provides only 4 metadata entries. This mismatch can cause UI rendering issues or missing icons for the Ollama provider.

Additionally, the metadata at line 1676 reuses the "OpenAI" icon for what appears to be IBM watsonx.ai. Provide metadata for all 5 providers with appropriate icons.

Apply this fix to align metadata with options:

                "options": [
                  "Anthropic",
                  "Google Generative AI",
                  "OpenAI",
                  "IBM watsonx.ai",
                  "Ollama"
                ],
                "options_metadata": [
                  {
                    "icon": "Anthropic"
                  },
                  {
                    "icon": "GoogleGenerativeAI"
                  },
                  {
                    "icon": "OpenAI"
                  },
-                  {
+                  {
-                    "icon": "OpenAI"
+                    "icon": "IBM"
+                  },
+                  {
+                    "icon": "Ollama"
                  }
                ],

src/lfx/src/lfx/base/models/model_input_constants.py (1)

24-57: Don’t disable model_name refresh for Ollama

Ollama relies on real_time_refresh/refresh_button to populate models. Current logic disables it for all except IBM.

Fix:

-    elif component_data.name == "model_name":
-        if provider_name not in ["IBM watsonx.ai"]:
-            component_data = set_real_time_refresh_false(component_data)
+    elif component_data.name == "model_name":
+        if provider_name not in ["IBM watsonx.ai", "Ollama"]:
+            component_data = set_real_time_refresh_false(component_data)
         component_data = add_combobox_true(component_data)

Alternatively, don’t override real_time_refresh at all and let components define it.

src/backend/base/langflow/initial_setup/starter_projects/Pokédex Agent.json (1)

1347-1364: [Duplicate from SaaS Pricing] Provider options expanded without corresponding metadata entries.

The agent_llm options list (lines 1347–1353) includes 5 providers ("Anthropic", "Google Generative AI", "OpenAI", "IBM watsonx.ai", "Ollama"), but options_metadata (lines 1354–1364) contains only 3 entries. This identical mismatch as in the SaaS Pricing starter project will cause rendering or indexing issues for the new providers.

Add metadata entries for IBM watsonx.ai and Ollama to maintain consistency.

src/backend/base/langflow/initial_setup/starter_projects/SaaS Pricing.json (1)

976-993: Provider options expanded without corresponding metadata entries.

The agent_llm options list (lines 976–982) includes 5 providers ("Anthropic", "Google Generative AI", "OpenAI", "IBM watsonx.ai", "Ollama"), but options_metadata (lines 983–993) contains only 3 entries with icons for Anthropic, GoogleGenerativeAI, and OpenAI.

Add metadata entries for "IBM watsonx.ai" and "Ollama" with their corresponding icons:

{
  "icon": "IBMWatsonx"
},
{
  "icon": "Ollama"
}

🧹 Nitpick comments (9)

src/backend/base/langflow/initial_setup/starter_projects/Youtube Analysis.json (1)

949-950: Strip the leftover debug print

print(watsonx_inputs_filtered) will spam stdout every time the component is imported. Please remove the print (or replace it with structured logging if you genuinely need the signal) before shipping.

src/backend/base/langflow/initial_setup/starter_projects/Search agent.json (1)

1135-1136: Remove the debug print

This starter carries the same print(watsonx_inputs_filtered) side effect. Please delete it or replace it with proper logging before release.

src/backend/base/langflow/initial_setup/starter_projects/Invoice Summarizer.json (1)

1265-1268: Provider options extended — add icons for new entries

IBM watsonx.ai and Ollama added, but options_metadata lacks their icons; add for visual consistency.

src/lfx/src/lfx/components/ibm/watsonx.py (1)

24-31: Static endpoints OK; allow custom endpoint too

List is fine; consider combobox True on base_url to allow custom regions/private gateways.

src/lfx/src/lfx/base/models/model_input_constants.py (2)

17-22: Style nit

Add space after commas in function signatures for readability.

---

355-360: Dynamic fields alignment

MODEL_DYNAMIC_UPDATE_FIELDS includes "watsonx_endpoint" and "url", while components use "base_url". Align names to avoid dead updates.

src/backend/base/langflow/initial_setup/starter_projects/News Aggregator.json (1)

1289-1292: Provider options extended — add icons for new entries

Add icons metadata for IBM watsonx.ai and Ollama to match others.

src/backend/base/langflow/initial_setup/starter_projects/SaaS Pricing.json (1)

1023-1041: Required fields with show: false depend on conditional display logic.

Lines 1023–1041 (base_url) and 1386–1404 (project_id) are both marked as required: true but show: false. These fields must become visible and enabled when users select IBM watsonx.ai or Ollama from the provider dropdown. This visibility toggling is likely handled by the update_build_config method in the embedded Python code.

Recommendation: Verify that the update_build_config method (or equivalent logic in the Agent component) correctly sets show: true for these fields when the corresponding provider is selected, otherwise users will encounter validation errors on submission.

Also applies to: 1386-1404

src/backend/base/langflow/initial_setup/starter_projects/Pokédex Agent.json (1)

1394-1412: [Duplicate from SaaS Pricing] Required fields with show: false depend on conditional display logic.

Lines 1394–1412 (base_url) and 1757–1775 (project_id) are marked as required: true but show: false, replicating the same pattern as SaaS Pricing. Ensure the Agent component's update_build_config method correctly toggles visibility when IBM watsonx.ai or Ollama providers are selected.

Also applies to: 1757-1775

📜 Review details

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between fa21c4e and c805135.

📒 Files selected for processing (19)

• src/backend/base/langflow/initial_setup/starter_projects/Instagram Copywriter.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Invoice Summarizer.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Knowledge Ingestion.json (2 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Knowledge Retrieval.json (2 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Market Research.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/News Aggregator.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Nvidia Remix.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Pokédex Agent.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Price Deal Finder.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Research Agent.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/SaaS Pricing.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Search agent.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Simple Agent.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Social Media Agent.json (6 hunks)
• src/backend/base/langflow/initial_setup/starter_projects/Youtube Analysis.json (6 hunks)
• src/lfx/src/lfx/base/models/model_input_constants.py (5 hunks)
• src/lfx/src/lfx/components/agents/agent.py (4 hunks)
• src/lfx/src/lfx/components/ibm/watsonx.py (4 hunks)
• src/lfx/src/lfx/components/ollama/ollama.py (2 hunks)

🧰 Additional context used

🧬 Code graph analysis (4)

src/lfx/src/lfx/components/ollama/ollama.py (1)

src/lfx/src/lfx/inputs/inputs.py (1)

• BoolInput (414-426)

src/lfx/src/lfx/components/ibm/watsonx.py (2)

src/lfx/src/lfx/custom/custom_component/component.py (1)

• get_base_inputs (167-170)

src/lfx/src/lfx/inputs/inputs.py (1)

• DropdownInput (465-490)

src/lfx/src/lfx/components/agents/agent.py (1)

src/lfx/src/lfx/inputs/inputs.py (3)

• SecretStrInput (286-341)
• StrInput (126-182)
• IntInput (344-376)

src/lfx/src/lfx/base/models/model_input_constants.py (3)

src/lfx/src/lfx/custom/custom_component/component.py (1)

• get_base_inputs (167-170)

src/lfx/src/lfx/components/ollama/ollama.py (1)

• ChatOllamaComponent (18-327)

src/lfx/src/lfx/components/ibm/watsonx.py (1)

• WatsonxAIComponent (16-207)

🔇 Additional comments (20)

src/backend/base/langflow/initial_setup/starter_projects/Knowledge Retrieval.json (2)

514-514: ✓ code_hash update is appropriate.

The metadata hash has been updated to reflect the refactored implementation with lazy-loaded path resolution. This ensures the component metadata accurately represents the current code state.

---

608-609: ✓ Lazy-loading refactor is well-executed.

The implementation correctly defers knowledge bases root path initialization:

• Global cache variable _KNOWLEDGE_BASES_ROOT_PATH is initialized on first use
• _get_knowledge_bases_root_path() safely guards initialization with a None check and proper global declaration
• Both update_build_config() and retrieve_data() call the getter instead of accessing the path directly
• Error handling is preserved (ValueError when knowledge directory is not set)
• Path expansion with .expanduser() handles home directory references appropriately

The refactor improves module startup performance by deferring settings resolution to first use, which is particularly valuable for components that may not always be instantiated.

src/backend/base/langflow/initial_setup/starter_projects/Knowledge Ingestion.json (3)

740-740: Approved: Lazy-loaded knowledge base root path refactoring.

The introduction of _get_knowledge_bases_root_path() and _get_kb_root() improves initialization timing and reduces startup overhead by deferring path resolution until actually needed. The caching pattern in _cached_kb_path is well-implemented.

Also applies to: 869-870

---

869-870: Consider thread-safety implications of global mutable state.

The module-level _KNOWLEDGE_BASES_ROOT_PATH global is mutated during lazy initialization. While the pattern is common and includes a None-check guard, in concurrent scenarios (e.g., multiple threads calling _get_knowledge_bases_root_path() simultaneously), a race condition could occur where the path is computed multiple times. If this component runs in a multi-threaded context, consider using thread-safe initialization (e.g., threading.Lock or threading.local()).

Does the Langflow backend execute component code in a multi-threaded environment where concurrent access to _get_knowledge_bases_root_path() is possible?

---

740-740: Verify file alignment with PR objectives.

The PR objectives reference updating "Instagram Copywriter.json" with new Agent inputs (base_url, project_id, max_output_tokens) and expanded provider support (IBM watsonx.ai, Ollama). However, the file under review is "Knowledge Ingestion.json" with knowledge base refactoring changes.

Please confirm whether this is the intended file for review or if "Instagram Copywriter.json" should be provided instead.

src/lfx/src/lfx/components/ollama/ollama.py (2)

194-194: LGTM! Defensive fallback added.

The parameter mapping correctly uses the new field name enable_verbose_output while maintaining the underlying library parameter name verbose. The or False fallback ensures a valid boolean value is always passed.

---

115-115: <!-- [scratchpad_update]
[task verify_breaking_change_impact 3/3 - FINAL]
Web search provided context about Ollama's verbose usage but no specific LFX/Langflow migration info. Result 9-1 shows LangChain's Ollama class has a "verbose" property. However, this is LangChain's verbose (debugging), not an input parameter name.

Key finding: The review comment treats this as a critical breaking change but there is:

1. NO evidence of existing "verbose" field in LFX Ollama component historically
2. NO migration path mentioned anywhere
3. Component input is BoolInput with name="enable_verbose_output" - this is the ONLY reference found
4. This appears to be either: (a) new component, (b) preemptive refactoring for clarity before release, or (c) misunderstanding in original review

Conclusion: The original review comment is likely INCORRECT or based on outdated information. The field name "enable_verbose_output" appears to be the current/only state. No evidence of previous "verbose" field or migration path needed.

[verdict]: Original review comment overstates the issue. This is NOT a breaking change for existing flows because:

• No historical "verbose" field found in LFX codebase
• This is a local component input rename
• No public API impact unless flows were serialized with old name
• Descriptive naming is actually a positive improvement
  [scratchpad_end] -->

No migration path needed - field name appears to be current state only.

Based on verification of the codebase and web search results, the original review comment appears to overstate the concern. The Ollama API does support a "verbose" parameter, but there is no evidence that the LFX Ollama component previously had a verbose field that was renamed to enable_verbose_output.

The script output confirms line 115 shows:

BoolInput(name="enable_verbose_output", display_name="Ollama Verbose Output", info="Whether to print out response text.", advanced=True),

This is the only field name found in the codebase. No verbose field, deprecated aliases, or migration logic exists in the component. The more descriptive field name enable_verbose_output is actually a UX improvement over a generic verbose name, especially since verbose is context-dependent (LangChain also uses verbose for debug logging).

There are no existing flows using an old verbose field name that would be affected by this change.

Likely an incorrect or invalid review comment.

src/lfx/src/lfx/components/agents/agent.py (2)

499-503: LGTM! Defensive programming enhancement.

The addition of build_config is not None check prevents potential AttributeError if delete_fields is called with a None build_config. This is good defensive programming.

---

23-23: LGTM! Required imports added.

The imports of SecretStrInput and StrInput are necessary for the new input fields added to the AgentComponent.

src/backend/base/langflow/initial_setup/starter_projects/Simple Agent.json (1)

1038-1044: Provider registration verified.

Both "IBM watsonx.ai" and "Ollama" are properly registered in MODEL_PROVIDERS_DICT with configuration at lines 321 and 335 respectively in model_input_constants.py, and both appear in MODEL_PROVIDERS_LIST. The corresponding components exist with test coverage. No further action required.

src/backend/base/langflow/initial_setup/starter_projects/Instagram Copywriter.json (2)

2137-2155: Verify conditional visibility logic for provider-specific fields.

Three new fields—base_url (line 2137), project_id (line 2500), and max_output_tokens (line 2281)—are configured with "show": false. These appear intended for conditional display based on the selected provider (e.g., only show for IBM watsonx.ai or Ollama). Ensure that the component's update_build_config method correctly sets show: true for these fields when the corresponding provider is selected.

Verify that the component logic in the Agent class (lines 2189–2645) conditionally updates these field visibilities. Look for code that toggles the show property based on self.agent_llm provider selection. If this logic is missing, the fields will remain hidden and users won't be able to configure them.

Also applies to: 2281-2298, 2500-2518

---

1966-1966: Confirm code_hash update reflects all component changes.

The code_hash metadata was updated from bdb8b8db6375 to 6f2517926f3d (line 1966). This reflects the changes to the Agent component inputs and code logic. Verify this hash is regenerated from the actual component code to avoid inconsistencies that could cause version mismatch issues in deployment.

src/backend/base/langflow/initial_setup/starter_projects/Price Deal Finder.json (1)

1714-1741: Add missing options_metadata entries using repository icon aliases.

Options include "IBM watsonx.ai" and "Ollama" but options_metadata only lists three icons; add the two missing entries and use the repo's alias for IBM watsonx.ai ("WatsonxAI"):

                 "options_metadata": [
                   {
                     "icon": "Anthropic"
                   },
                   {
                     "icon": "GoogleGenerativeAI"
                   },
                   {
                     "icon": "OpenAI"
                   }
+                  {
+                    "icon": "WatsonxAI"
+                  },
+                  {
+                    "icon": "Ollama"
+                  }
                 ],

Likely an incorrect or invalid review comment.

src/backend/base/langflow/initial_setup/starter_projects/Invoice Summarizer.json (2)

1154-1154: Metadata bump OK

code_hash update acknowledged.

---

1453-1470: IntInput default as empty string — confirm type handling

max_output_tokens has value:"" for an int field. Ensure renderer/parsing tolerates empty string; otherwise prefer null/omitted.

src/lfx/src/lfx/components/ibm/watsonx.py (2)

60-63: Model Name refresh wiring looks good

real_time_refresh + refresh_button for model_name is appropriate.

---

200-207: build_model will receive list if base_url remains list

This is fixed by the base_url value change above; otherwise ChatWatsonx(...) will break at runtime.

src/lfx/src/lfx/base/models/model_input_constants.py (1)

318-345: Provider registrations look good

IBM watsonx.ai and Ollama entries added correctly with icons and active flags.

src/backend/base/langflow/initial_setup/starter_projects/News Aggregator.json (2)

1178-1178: Metadata bump OK

code_hash update acknowledged.

---

1477-1494: IntInput default as empty string — confirm type handling

max_output_tokens has value:"" for an int field. Verify UI/parser accepts empty string.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

Remove debug print statement from embedded code.

Line 2172 contains a debug print(watsonx_inputs_filtered) statement in the component code. This should be removed before release.

Apply this diff:

    if \"IBM watsonx.ai\" in MODEL_PROVIDERS_DICT:
        watsonx_inputs_filtered = list(MODEL_PROVIDERS_DICT[\"IBM watsonx.ai\"][\"inputs\"])
-       print(watsonx_inputs_filtered)
    else:
        openai_inputs_filtered = []

🤖 Prompt for AI Agents

src/backend/base/langflow/initial_setup/starter_projects/Instagram
Copywriter.json around line 2172: the embedded Python code contains a stray
debug print statement print(watsonx_inputs_filtered) which should be removed;
edit the file to delete that print call (ensuring surrounding logic and variable
assignment remain intact) so no debug output is emitted from the component.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Hidden + required field will block runs

base_url is required:true but show:false and default:"". This can fail validation when provider ≠ IBM/Ollama.

Recommended change (make non-required at base level; provider components will enforce when selected):

 "base_url": {
-  "required": true,
+  "required": false,
   "show": false,
   "value": ""
 }

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	"base_url": {
	"_input_type": "StrInput",
	"advanced": false,
	"display_name": "Base URL",
	"dynamic": false,
	"info": "The base URL of the API.",
	"list": false,
	"list_add_label": "Add More",
	"load_from_db": false,
	"name": "base_url",
	"placeholder": "",
	"required": true,
	"show": false,
	"title_case": false,
	"tool_mode": false,
	"trace_as_metadata": true,
	"type": "str",
	"value": ""
	},
	"base_url": {
	"_input_type": "StrInput",
	"advanced": false,
	"display_name": "Base URL",
	"dynamic": false,
	"info": "The base URL of the API.",
	"list": false,
	"list_add_label": "Add More",
	"load_from_db": false,
	"name": "base_url",
	"placeholder": "",
	"required": false,
	"show": false,
	"title_case": false,
	"tool_mode": false,
	"trace_as_metadata": true,
	"type": "str",
	"value": ""
	},

🤖 Prompt for AI Agents

In src/backend/base/langflow/initial_setup/starter_projects/Invoice
Summarizer.json around lines 1309 to 1327, the base_url field is marked
required:true while show:false and value:"", which will block runs when
providers other than IBM/Ollama are used; change the base-level field to
required:false (so it no longer enforces validation globally) and keep
provider-specific components responsible for marking base_url required when
those providers are selected.

[coderabbitai]

🛠️ Refactor suggestion | 🟠 Major

Agent code: stray print and logic bug wiping OpenAI inputs

• print(watsonx_inputs_filtered) is a debug artifact.
• Else branch tied to IBM check resets openai_inputs_filtered = [] and can remove OpenAI inputs unexpectedly.

Apply inside the code string:

-    if "OpenAI" in MODEL_PROVIDERS_DICT:
-        openai_inputs_filtered = [
-            input_field
-            for input_field in MODEL_PROVIDERS_DICT["OpenAI"]["inputs"]
-            if not (hasattr(input_field, "name") and input_field.name == "json_mode")
-        ]
-    if "IBM watsonx.ai" in MODEL_PROVIDERS_DICT:
-        watsonx_inputs_filtered = list(MODEL_PROVIDERS_DICT["IBM watsonx.ai"]["inputs"])
-        print(watsonx_inputs_filtered)
-    else:
-        openai_inputs_filtered = []
+    openai_inputs_filtered = []
+    if "OpenAI" in MODEL_PROVIDERS_DICT:
+        openai_inputs_filtered = [
+            input_field
+            for input_field in MODEL_PROVIDERS_DICT["OpenAI"]["inputs"]
+            if not (hasattr(input_field, "name") and input_field.name == "json_mode")
+        ]
+    if "IBM watsonx.ai" in MODEL_PROVIDERS_DICT:
+        watsonx_inputs_filtered = list(MODEL_PROVIDERS_DICT["IBM watsonx.ai"]["inputs"])

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Required fields should not be hidden by default.

The base_url field is marked as required: true but show: false. This creates a UX issue—users cannot fulfill a required field that is hidden. If this field should only be required for certain providers, use conditional visibility in the update_build_config method or set required: false if optional.

Consider adjusting the configuration:

              "base_url": {
                "_input_type": "StrInput",
                "advanced": false,
                "display_name": "Base URL",
                "dynamic": false,
                "info": "The base URL of the API.",
                "list": false,
                "list_add_label": "Add More",
                "load_from_db": false,
                "name": "base_url",
                "placeholder": "",
-               "required": true,
+               "required": false,
                "show": false,

🤖 Prompt for AI Agents

In src/backend/base/langflow/initial_setup/starter_projects/Market Research.json
around lines 1708 to 1726, the base_url property is marked required: true but
show: false, which prevents users from entering a required value; change this by
either setting "show": true so the field is visible when required, or set
"required": false if it is truly optional, or implement conditional visibility
in update_build_config so base_url is only required (required: true and show:
true) for providers that need it and hidden/optional for others—apply one of
these fixes and keep metadata (display_name/info) consistent.

[coderabbitai]

🛠️ Refactor suggestion | 🟠 Major

Implement provider-specific conditional visibility for new fields.

The new fields base_url, project_id, and max_output_tokens are added as universal inputs, but they should be conditionally shown/required based on the selected provider. For example:

• base_url and project_id are specific to IBM watsonx.ai
• base_url may also be relevant for Ollama
• max_output_tokens is provider-specific

The update_build_config method should dynamically show/hide and mark these fields as required based on the selected provider. Consider this pattern:

async def update_build_config(self, build_config: dotdict, field_value: str, field_name: str | None = None) -> dotdict:
    if field_name == "agent_llm":
        # Hide all provider-specific fields by default
        build_config["base_url"]["show"] = False
        build_config["base_url"]["required"] = False
        build_config["project_id"]["show"] = False
        build_config["project_id"]["required"] = False
        build_config["max_output_tokens"]["show"] = False
        
        # Show/require fields based on selected provider
        if field_value == "IBM watsonx.ai":
            build_config["base_url"]["show"] = True
            build_config["base_url"]["required"] = True
            build_config["project_id"]["show"] = True
            build_config["project_id"]["required"] = True
            build_config["max_output_tokens"]["show"] = True
        elif field_value == "Ollama":
            build_config["base_url"]["show"] = True
            # Ollama has default, so not required
            build_config["base_url"]["required"] = False
    
    # ... rest of the method

Also update the initial field definitions to have required=False:

  StrInput(
      name="base_url",
      display_name="Base URL",
      info="The base URL of the API.",
-     required=True,
+     required=False,
      show=False,
  ),
  StrInput(
      name="project_id",
      display_name="Project ID",
      info="The project ID of the model.",
-     required=True,
+     required=False,
      show=False,
  ),

🤖 Prompt for AI Agents

In src/lfx/src/lfx/components/agents/agent.py around lines 83 to 106, the new
inputs base_url, project_id, and max_output_tokens are always visible/required
but must be provider-specific; change their initial definitions to
required=False, and update the update_build_config method to, when field_name ==
"agent_llm", default-hide and mark not required for these fields then
conditionally set show=True and required=True as appropriate for each provider
(e.g., IBM watsonx.ai: show+require base_url and project_id and show
max_output_tokens; Ollama: show base_url but keep required=False), leaving other
providers hidden.

[edwinjosechittilappilly]

why required=False?

[coderabbitai]

@edwinjosechittilappilly Good question! The reason for required=False as the initial/default value is because these fields are provider-specific:

• base_url and project_id are only needed for IBM watsonx.ai (and base_url for Ollama)
• max_output_tokens is only relevant for certain providers

If we keep required=True as the default, then when a user selects OpenAI or Anthropic (which don't need these fields), they would still be marked as required and the validation would force users to provide values even though those providers don't use them.

The pattern is:

1. Start with required=False (universal safe default for all providers)
2. In update_build_config, when field_name == "agent_llm", dynamically set required=True for the specific providers that actually need those fields

This way:

• IBM watsonx.ai selected → base_url and project_id become required=True and show=True
• Ollama selected → base_url becomes show=True but stays required=False (has a default)
• OpenAI selected → these fields stay required=False and show=False (hidden and not validated)

This ensures proper validation only when the fields are relevant to the selected provider! 🐰

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Clarify incomplete Watsonx integration.

The Watsonx inputs are filtered (lines 54-56) but never used in the inputs list (line 108 is commented out). This suggests incomplete implementation.

Either:

1. Complete the Watsonx integration by uncommenting this line and ensuring proper provider-specific field handling, or
2. Remove the unused filtering code (lines 54-56) and add a TODO comment for future implementation, or
3. Document why this is prepared but not activated

  *openai_inputs_filtered,
- # *watsonx_inputs_filtered,
+ # TODO: Watsonx inputs integration pending provider-specific field handling
  MultilineInput(

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	# *watsonx_inputs_filtered,
	*openai_inputs_filtered,
	# TODO: Watsonx inputs integration pending provider-specific field handling
	MultilineInput(

🤖 Prompt for AI Agents

In src/lfx/src/lfx/components/agents/agent.py around line 108 (and referencing
the filtering at lines 54-56), the watsonx_inputs are filtered but never used
(the inclusion at line 108 is commented out), indicating incomplete integration;
fix by either (A) re-enable watsonx inputs by uncommenting that line and ensure
provider-specific fields from the filtered watsonx_inputs are correctly mapped
into the inputs list and validated for watsonx expectations, or (B) remove the
unused filtering code at lines 54-56 and add a clear TODO comment at line 108
noting that watsonx support is pending, or (C) leave the filtering but add a
concise explanatory comment at both locations documenting why watsonx inputs are
prepared but intentionally not included, then run tests to ensure no
unused-variable or lint warnings remain.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

base_url type/value mismatch (list vs str) and UX

Dropdown value=[] will pass a list into build_model; ChatWatsonx.url expects a string. Also no default selected.

Fix:

 DropdownInput(
-    name="base_url",
+    name="base_url",
     display_name="watsonx API Endpoint",
     info="The base URL of the API.",
-    value=[],
-    options=_urls,
+    value=_urls[0],
+    options=_urls,
+    combobox=True,
     real_time_refresh=True,
     required=True,
 ),

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	name="base_url",
	display_name="watsonx API Endpoint",
	info="The base URL of the API.",
	value=None,
	options=[
	"https://us-south.ml.cloud.ibm.com",
	"https://eu-de.ml.cloud.ibm.com",
	"https://eu-gb.ml.cloud.ibm.com",
	"https://au-syd.ml.cloud.ibm.com",
	"https://jp-tok.ml.cloud.ibm.com",
	"https://ca-tor.ml.cloud.ibm.com",
	],
	value=[],
	options=_urls,
	real_time_refresh=True,
	required=True,
	),
	name="base_url",
	display_name="watsonx API Endpoint",
	info="The base URL of the API.",
	value=_urls[0],
	options=_urls,
	combobox=True,
	real_time_refresh=True,
	required=True,
	),

🤖 Prompt for AI Agents

In src/lfx/src/lfx/components/ibm/watsonx.py around lines 35 to 42, the
"base_url" field sets value=[] (a list) but ChatWatsonx.url expects a string and
there is no default selected; change the field to provide a string value that
matches the options (e.g., set value to the first entry of _urls or an empty
string ""), ensuring the value type is str and not a list, and confirm _urls
contains string URLs so the dropdown passes a single string into build_model.

[edwinjosechittilappilly]

By Default I do not need any values

[coderabbitai]

@edwinjosechittilappilly, understood! If you don't want a pre-selected default, that's perfectly fine since the field is marked required=True and users will need to select a value before building.

One small note: for a dropdown with no default, the value should typically be None or "" (empty string) rather than [] (empty list) to maintain type consistency, since self.base_url will eventually be passed as a string to ChatWatsonx(url=self.base_url, ...).

If the current implementation with value=[] works correctly in your validation flow, then no changes are needed. Just wanted to flag the type consideration in case it causes any unexpected behavior.

[coderabbitai]

⚠️ Potential issue | 🔴 Critical

update_build_config: wrong condition and defaulting

• Ties model_name updates to field_name=="model_name" and checks value in _urls; should key off base_url.
• Defaulting only when value is truthy skips setting a valid default.

Apply:

-        if field_name == "base_url" and field_value:
+        if field_name == "base_url" and field_value:
             try:
                 models = self.fetch_models(base_url=field_value)
-                build_config["model_name"]["options"] = models
-                if build_config["model_name"]["value"]:
-                    build_config["model_name"]["value"] = models[0]
+                build_config["model_name"]["options"] = models
+                current = build_config["model_name"].get("value")
+                build_config["model_name"]["value"] = models[0] if not current or current not in models else current
                 info_message = f"Updated model options: {len(models)} models found in {field_value}"
                 logger.info(info_message)
             except Exception:  # noqa: BLE001
                 logger.exception("Error updating model options.")
-        if field_name == "model_name" and field_value and field_value in WatsonxAIComponent._urls:
-            build_config["model_name"]["options"] = self.fetch_models(base_url=field_value)
-            build_config["model_name"]["value"] = ""
+        # No-op for model_name here; refresh is driven by base_url changes.
         return build_config

🤖 Prompt for AI Agents

In src/lfx/src/lfx/components/ibm/watsonx.py around lines 160 to 173, the second
conditional incorrectly keys off field_name == "model_name" and only sets a
default when the incoming value is truthy; change that conditional to check the
base_url (e.g., if field_name == "base_url" and the base_url value exists in
WatsonxAIComponent._urls), fetch model options for that base_url and assign them
to build_config["model_name"]["options"], and always set a sensible default for
build_config["model_name"]["value"] (preferably models[0] when models exist, or
"" if none) instead of only setting it when a truthy value is provided. Ensure
you reuse/avoid duplicate fetches if the earlier base_url branch already covered
this, and keep exception handling around the fetch as in the first block.

[edwinjosechittilappilly]

I tried the earlier approach but this seems to work with agents well.

[coderabbitai]

@edwinjosechittilappilly, thank you for the clarification! I appreciate that you've tested this approach and confirmed it works well with agents. Since you have practical experience showing this implementation is effective, that's what matters most.

The second conditional (checking field_name == "model_name" with field_value in _urls) is a bit non-obvious at first glance, so you might consider adding a brief comment explaining the scenario it handles (e.g., "Handle refresh when model_name receives a URL value from the refresh button" or similar). But that's entirely optional—if it's working as expected, you're good to go!

Feel free to resolve this thread when you're ready. 🐰

[lucaseduoli]

lgtm