Swarms Contributors Portal

1515

Rank #1

elder-plinius

Contributor

258

Rank #2

dependabot[bot]

Bot

200

Rank #3

IlumCI

Contributor

174

Rank #4

harshalmore31

Contributor

132

Rank #5

evelynmitchell

Contributor

119

Rank #6

Wyatt-Stanke

Contributor

102

Rank #7

Steve-Dusty

Contributor

Rank #8

ascender1729

Contributor

Rank #9

adichaudhary

Contributor

Rank #10

aparekh02

Contributor

Rank #11

hughiwnl

Contributor

Rank #12

sambhavnoobcoder

Contributor

Rank #13

ZackBradshaw

Contributor

Rank #14

ayaangazali

Contributor

Rank #15

GoldenWind8

Contributor

Rank #16

Wxysnx

Contributor

Rank #17

nathan-d

Contributor

Rank #18

vyomakesh09

Contributor

Rank #19

Occupying-Mars

Contributor

Rank #20

patrickbdevaney

Contributor

Rank #21

w-zz-w-zz-w

Contributor

Rank #22

filip-michalsky

Contributor

Rank #23

chethanuk

Contributor

Rank #24

gilbertoaceville

Contributor

Rank #25

nicorne

Contributor

Rank #26

eltociear

Contributor

Rank #27

Fadope1

Contributor

Rank #28

adtygan

Contributor

Rank #29

thecaptain789

Contributor

Rank #30

seigot

Contributor

Rank #31

cyberholics

Contributor

Rank #32

nullonesix

Contributor

Rank #33

exvul-nolan

Contributor

Rank #34

nictuku

Contributor

Rank #35

kuant

Contributor

Rank #36

krankos

Contributor

Rank #37

idreesaziz

Contributor

Rank #38

grit-app[bot]

Bot

Rank #39

argentpapa

Contributor

Rank #40

RinZ27

Contributor

Rank #41

RahulK4102

Contributor

Rank #42

Nishant-k-sagar

Contributor

Rank #43

MycCellium420

Contributor

Rank #44

MalikHaroonKhokhar

Contributor

Rank #45

lidashu

Contributor

Rank #46

cursoragent

Contributor

Rank #47

bookofbash

Contributor

Rank #48

AlexanderLavelle

Contributor

Upper9527•Opened 16 minutes ago

Rank #49

Active Tasks

#1872 Possible complementary direction: LongHorizon-Harness for sustained agent tasks

#1864 [Bug][RoundRobinSwarm] self.index is written but never read — run_batch gives agents[0] the opening turn on every task

kyegomez•Opened 1 day ago

#1853 [Bug] Six smaller correctness defects found during the code-waste audit

#1852 [Bug][Agent] agent.run() with an empty task blocks on stdin even when interactive=False — a commented-out guard inverted the feature

#1851 [Cleanup] Four overlapping 'format any data structure as a string' helpers (~80 lines) — keep one

#1850 [Cleanup] Introduce SwarmMixin — 22 batch_run copies, 13 call copies, 13 reliability_check copies, 8 hand-rolled thread pools (~850 lines)

#1849 [Cleanup] Ten smaller confirmed clones across structs/tools/utils (~605 lines)

#1848 [Bug][Cleanup] Pydantic→OpenAI schema conversion exists 3 times, round-trips through a legacy envelope, and emits 'ModelMetaclass' as the function name

#1847 [Cleanup] The same OpenAI tool_call parser is written 8+ times across base_tool and 3 swarms (~240 lines)

#1846 [Cleanup] graph_workflow.py — 4 edge methods share one body, validate duplicates _fast_validate, 21 no-op try/except wrappers (~490 lines)

#1845 [Cleanup] agent.py _run_autonomous_loop — 862 lines in one method, 15 levels deep, with the dispatch block written twice (~250 lines)

#1844 [Cleanup][AOP] MCP tool closures re-wrap methods defined 30 lines above (~445 lines), plus redundant re-registration and an unreachable branch

#1843 [Cleanup] Autosave pair copy-pasted into 5 orchestrators (~335 lines) — extract a SwarmAutosaveMixin

#1842 [Cleanup] heavy_swarm.py — two copies of the executor and four copies of the variant table (~350 lines)

#1841 [Cleanup] multi_agent_debates.py — one class template pasted seven times (~470 lines)

#1840 [Cleanup] conversation.py — 6 dead methods, duplicate accessors, and a broken save/load round trip (~190 lines)

#1839 [Bug][GraphWorkflow] to_json/from_json produce unrunnable workflows — delete the broken second serialization stack (495 lines)

#1838 [Cleanup] formatter.py — 3 dead print methods (~142 lines) and 8 collapsible log-strip regexes

#1837 [Cleanup] 12 unreferenced functions scattered across swarms/structs/ (~90 lines) — keep/drop decision needed

question

#1836 [Cleanup] agent.py — 15 more dead methods (~240 lines) with zero call sites

#1835 [Cleanup] Delete 23 unused prompt files in swarms/prompts/ (~1,677 lines)

prompts

#1834 [Cleanup] Remaining dead modules (tool_registry, collaborative_utils) and ~160 lines of in-file dead code

#1833 [Cleanup] base_tool.py — 17 never-called public methods (~1,060 lines); only 4 of 30 are used

#1832 [Cleanup] base_tool.py:1567-2193 — delete the 627-line schema validation/conversion island (zero callers)

#1831 [Cleanup] Delete the remaining 527 lines of various_alt_swarms.py — dead duplicate of swarming_architectures.py

#1824 [Cleanup] base_swarm.py: remove 31 empty-body stubs + 19 unreferenced methods (~300 lines) and migrate its one subclass

#1798 [Bug][CI][Four workflow defects: hardcoded local macOS path, unscoped pytest, secretless unbounded test job, and a Python 3.9 release build]

#1794 [Bug][Agent][tool_execution_retry never retries and silently swallows tool failures]

#1793 [Bug][Agent][self.executor is never assigned in init: run(imgs=[...]) raises AttributeError]

#1791 [Bug][Autonomous loop][File tools accept ../ and absolute paths, so the agent can read, write and delete anywhere on the filesystem]

#1788 [Bug][multi_agent_exec][run_agents_concurrently returns completion order but MajorityVoting, AgentRearrange and aggregate zip it against input order]

Steve-Dusty•Opened 11 days ago

#1781 [Feature][MCP][Support mcp 2.x: migrate AOP off the removed FastMCP and rename the streamable-HTTP client]

#1778 [Improvement][SocialAlgorithms][Add OpenTelemetry instrumentation to both the sync and async run paths]

#1777 [Improvement][AutoSwarmBuilder][Add OpenTelemetry instrumentation so the build phase and delegated run share one trace]

#1776 [Improvement][SelfMoASeq][Add OpenTelemetry instrumentation to the sequential mixture-of-agents loop]

#1775 [Improvement][SpreadSheetSwarm][Add OpenTelemetry instrumentation to the swarm-level run path]

#1774 [Improvement][ModelRouter][Add OpenTelemetry instrumentation and fix trace-context loss across its thread pool]

#1758 [Improvement][GraphWorkflow][Per-node retry policy for transient agent failures]

#1757 [Bug][GraphWorkflow][Cycles silently flatten into one parallel layer]

[BUG]

#1756 [Feature][GraphWorkflow][Conditional edges for data-dependent routing]

FEAT

#1755 [Improvement][swarms.init][Lazy module loading to cut ~1.1s off cold start]

#1754 [Improvement][litellm and mcp are imported eagerly on `import swarms`][They account for ~58% of the 1.7s import time and neither is needed until first LLM call or first MCP use]

#1753 [Improvement][Autonomous loop front-loads all 16 built-in tool schemas on every call][~2,319 tokens of schema re-sent every iteration and never cached; load on demand via a search_tools meta-tool or a plan-time relevance pass]

#1752 [Improvement][Autonomous loop iteration limits are hardcoded module constants][MAX_PLANNING_ATTEMPTS, MAX_SUBTASK_ITERATIONS and MAX_SUBTASK_LOOPS should be Agent.init params]

#1751 [Feature][No permission layer for autonomous-loop file and bash tools][create_file, update_file, delete_file and run_bash execute unconditionally; the only guard is a 512-char length cap]

[SECURITY THREAT]

#1750 [Bug][Built-in autonomous-loop tools execute serially while user tools run concurrently][read_file, grep, list_directory and run_bash loop one at a time at agent.py:2402 despite base_tool supporting max_workers=4]

#1749 [Feature][Replace the mandatory plan gate in max_loops="auto" with a flat agentic loop][create_plan is forced and hard-raises on failure; the plan is frozen at t=0 with no path to revise after discovery]

#1748 [Bug][Autonomous loop flattens context into one string instead of a message array][Every iteration re-sends the whole transcript as a single user message, killing prompt caching and breaking tool_use_id pairing]

#1747 [Bug][Autonomous loop ships a redundant `think` tool by default][Gate keys off thinking_tokens, which is the wrong signal; the tool forces JSON-serialized reasoning and costs a full extra round-trip per call]

#1746 [Bug][dynamic_auto_chunking re-tokenizes the full transcript ~18x on every get_str() call][Any conversation history read above context_length costs 200-370ms of CPU; multiplied by agents x loops this is 22s for a 10x10 swarm]

memory

#1738 [Improvement] import swarms costs ~1.6s, ~86% of it eager litellm import

#1736 [Improvement] count_tokens is uncached and recomputed on the full growing history

#1735 [Improvement] Conversation history rebuilt and re-tokenized every turn — O(n²) in turn count

#1718 [BUG] `Agent._run` appends the response to `short_memory` before `success=True`, so a failing post-add step makes a retry double-commit the assistant message

IngridJiang•Opened 25 days ago

#1621 [improvement][Agent][stream thinking tokens through streaming_callback and arun_stream]