couchbaselabs / couchbaselabs/agent-catalog
agentc-llamaindex LlamaIndex version constraint blocks classic ReAct agent usage
- Dominant language
- Python
- Stars
- 18
- Forks
- 3
- Avg merge
- 2h 46m
- Merged PRs (30d)
- 2
Description
**Title:** agentc-llamaindex LlamaIndex version constraint blocks classic ReAct agent usage
---
## Problem
The `agentc-llamaindex` package currently requires:
```toml
llama-index = "^0.13.4"
```
This forces projects to use **LlamaIndex 0.13.x**, which introduced breaking API changes that removed the classic ReAct agent interface.
---
## API Changes in LlamaIndex 0.13.x
* **Removed:** `ReActAgent.from_tools()` method
* **Removed:** `.chat()` method on agents
* **Replaced with:** Workflow-based agent system with different API
* **Lost:** Classic verbose ReAct output format (`Thought: / Action: / Observation:`)
---
## Impact
* Cannot use classic ReAct agent with verbose debugging output
* Existing code using `ReActAgent.from_tools()` and `.chat()` breaks
* Development blocked due to dependency version solving failure
---
## Requested Solution
Relax the LlamaIndex version constraint in `agentc-llamaindex` from:
```toml
llama-index = "^0.13.4"
```
To:
```toml
llama-index = ">=0.12.0,<0.14.0"
```
This would allow projects to choose between:
* **LlamaIndex 0.12.x** → Classic ReAct agent API (verbose Thought/Action/Observation output)
* **LlamaIndex 0.13.x** → New workflow API
---
## Workaround
Currently using a temporary workaround by disabling `agentc-llamaindex` dependency, but this removes AgentC integration benefits.
---
## Environment
* **Python:** 3.12
* **agentc-llamaindex version:** 0.0.0
* **Desired LlamaIndex version:** 0.12.x for classic API compatibility
Contributor guide
Research direction
Start by locating the dependency declaration for the agentc-llamaindex package and compare its current LlamaIndex constraint with the requested range. Verify that dependency resolution permits both 0.12.x and 0.13.x, and confirm that the package remains installable with the supported Python environment.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ai
- Issue type
- Bug
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Stale
- Clarity
- Clearly specified
- Newbie friendliness
- 62/100