smolagents vs TradingAgents

smolagents 34

v1.26.0 on 29 May 2026, 161 days before this check (10). No release in the 90 days to 6 November (0). 416 open issues and 1,045 open issues and pull requests on 6 November, with 144 issues opened in the 90 days to 3 November. Of the 50 newest, 26 had a comment and 1 was closed. Two 2026 CVE records say the vendor didn't respond to the reporter, and seven commits reached main from 8 August to 6 November (6). Python only, and the package is the current one (10). CI with actions pinned to commit SHAs and Dependabot for actions (8). 10+0+6+10+8 = 34.

TradingAgents 78

0.7.1 on 27 October 2026 (30). Seven releases in the 90 days to 2 November, 0.4.0 (31 August), 0.5.0, 0.5.1, 0.5.2, 0.6.0, 0.7.0 and 0.7.1 (20). The earlier issue and pull request counts came from shields.io and are struck. On the clone, fixes cite issue numbers and each release credits the people who reported or fixed them (15 outside contributors in 0.7.1), but 251 of the 258 commits since 4 August are by one maintainer's accounts and reply times couldn't be read (13 of 25, as pipecat on clone evidence). No PyPI package; versions ship as GitHub releases with a wheel since 0.7.0 (8 of 15). CI with tests, lint and a clean-install check; dependencies set lower bounds only and no Dependabot configuration is in the repository (7 of 10). 30+20+13+8+7 = 78.

smolagents 81

Framework reading. A typed Python API with reference pages for agents, models, tools and executors (22). llms.txt and Markdown pages on huggingface.co (10). The docs say when to use CodeAgent or ToolCallingAgent and when a sandbox is needed (17). Tools declare typed inputs in a schema dictionary or from a decorated function (12). Many runnable examples; error classes exist but aren't catalogued (10). Versions tagged with release notes, no changelog file (10). 22+10+17+12+10+10 = 81.

TradingAgents 61

Framework reading. No API reference or docs site. Typed interfaces exist in the source, Pydantic models for the research manager's, trader's and portfolio manager's outputs and Annotated signatures on the twelve data tools, with no py.typed marker (10 of 25). The README and changelog are Markdown in the repository; no llms.txt (4 of 10). The README states limits plainly, that runs vary, which statements past-dated runs withhold and why, and that it is a research tool, not advice (15 of 20). Configuration is a plain dict of string keys, though environment overrides are coerced to the default's type and fail loudly on a bad value (8 of 15). Runnable Python and CLI examples throughout the README; the vendor error hierarchy is documented only in source docstrings (9 of 15). Dated changelog for every release from 0.1.0 with upgrade notes (15).

smolagents 64

Local-package reading. Official PyPI package with Python 3.10 or newer (20). A public tests workflow runs pytest on Python 3.10 and 3.12 for pushes and pull requests; whether main passes wasn't read (15). 416 open issues and 1,045 open issues and pull requests on 6 November, with 144 issues opened in the 90 days to 3 November. Of the 50 newest, 26 had a comment and 1 was closed. Two 2026 CVE records say the vendor didn't respond to the reporter, and seven commits reached main from 8 August to 6 November (6). Release notes on GitHub tags but no changelog file, and the remote WasmExecutor was removed in a minor release (8). 1.0 since 31 December 2024 (15). 20+15+6+8+15 = 64.

TradingAgents 52

Read with the local-software lines, because the owner runs it. Python 3.11 or later is stated and CI tests 3.11 to 3.14, but the project publishes no PyPI package. No workflow uploads one, and the README installs from a clone or Docker Compose. From 0.7.0 a release workflow builds a wheel and source archive from each tag and attaches them to the GitHub release (12 of 20). Public CI on every push and pull request runs the test suite (about 1,009 test functions at v0.7.1) on four Python versions and once in America/New_York, a clean-install import check and Ruff. The 25 October check read the pass state from shields.io, a third-party copy of GitHub data this run doesn't use, so that reading is struck, and the pass state on main is unread (15, as berth and agentvfs). Issue counts came from the same source and are struck too. On the clone, 0.7.1 (27 October) fixes 22 numbered issues, 17 of them above #1518, the newest fixed in 0.7.0 on 16 October, and credits 15 outside contributors; open crash or regression issues and reply times are unread (15, as berth on the same pattern). The changelog follows Keep a Changelog and Semantic Versioning and lists breaking changes under "Upgrading from", but 0.5.1, a patch release, moved import paths and removed classes (10 of 15). Version 0.7.1 with no stability statement (0). 12+15+15+10+0 = 52.

smolagents 68

Framework reading. MCPClient loads an MCP server's tools in a few lines (20). max_steps, planning intervals and final_answer_checks bound a run (16). Errors from tools and code go back to the model as observations (14). No built-in retry, idempotency or resume beyond replaying memory (8). Few required parameters and many providers, Python only, and the default model needs a Hugging Face token (10). 20+16+14+8+10 = 68.

TradingAgents 61

Framework reading, scored on what an outside agent needs to call it. No MCP client or server, so it can't be attached as tools; a run is four lines of Python (TradingAgentsGraph(config=...).propagate(ticker, date)) or one headless CLI command with flags (10 of 25). Analyst selection, debate and risk rounds (default 1), max_tool_rounds (20), max_recur_limit (100), article limits and --no-html size a run (14 of 20). A typed vendor error hierarchy separates unavailable from absent data, the headless CLI names the missing flag before a run starts, and an unparseable decision returns REVIEW rather than Hold (15 of 20). Checkpoint resume is opt-in, a backtest rerun with the same run_id skips finished cells, and settlement writes each decision once; nothing writes to an outside system (15 of 20). Sensible defaults, but the default provider is OpenAI's GPT-6 models with a key, and Python only (7 of 15).