1.1.0

What's Changed

• arxiv url by @xingdi-eric-yuan in #106
• Readme install from pypi by @matheper in #108
• Change credential order in AzureOpenAILLM for proper token retrieval by @matheper in #110
• AzureOpenAI Chained Credential by @matheper in #111
• Function Calling Syntax by @xingdi-eric-yuan in #109
• Update human class to use tool calls by @matheper in #114
• make sure the working dir is always in sys path by @xingdi-eric-yuan in #115
• Validate human input by @matheper in #116
• Bugfix human test by @matheper in #118
• Disable strict mode by @xingdi-eric-yuan in #117
• Fix env.rewrite_counter by @matheper in #119
• Fixed EvalTool being called with extra kwargs when reacting to events by @matheper in #120
• Add logs viewer for Froggy by @MarcCote in #95
• Fix: Use json.dumps for tool arguments in OpenAILLM by @ShiZhengyan in #123
• Tools observations by @matheper in #121
• Fix 'utf-8' codec error with surrogate pairs in Unicode strings by @Copilot in #129
• Add max_retries parameter to Human class to limit terminal read attempts by @Copilot in #126
• Improve test coverage report and default pytest configs by @matheper in #98
• Clean up pytest.ini, partially reverting #98 by @matheper in #132
• Remove current file by @matheper in #127
• Add start and end args to ViewTool by @matheper in #133
• Minor Fixes by @xingdi-eric-yuan in #124
• Fix kwargs by @xingdi-eric-yuan in #135
• Fix kwargs for pdb tool by @xingdi-eric-yuan in #136
• Fix type annotation: tool_call_list should be list not dict by @ShiZhengyan in #134
• Enhance Agent logging to include step number and reason for termination by @matheper in #137
• Improve visualization by @xingdi-eric-yuan in #138
• Pdb current frame file by @matheper in #139
• Pdb breakpoint handling by @matheper in #140
• Refactor llm_api into debug_gym.llms subpackage by @MarcCote in #142
• Resolve absolute path from RepoEnv by @matheper in #144
• Use better command completion for Human Mode by @MarcCote in #143
• Fix: resolve_path and is_editable to account for ignored and read-only files by @matheper in #145
• Ignore files from .gitignore by @matheper in #146
• Fix Aider ignore patterns and add tests for path resolution and ignored/read-only files by @matheper in #147
• Fix issue resolving env.working_dir by @matheper in #148
• Set default RepoEnv.dir_tree_depth to 1 by @matheper in #150
• replace unescape by filtering non-utf8 chars in system prompts by @xingdi-eric-yuan in #151
• Adding SWE-Smith support by @MarcCote in #122
• Only load image for instance_id we want to tests by @MarcCote in #154
• No eval shortcut by @matheper in #152
• Parallel execution by @matheper in #153
• Trajectory Filtering by @xingdi-eric-yuan in #141
• For SWE-Smith, add a new test split distinct from train-789 by @MarcCote in #156
• Fix ViewTool handling empty files by @matheper in #157
• Improve Retry by @xingdi-eric-yuan in #158
• Support pickling PDBTool instances by @threewisemonkeys-as in #166
• Resolve path mismatch issue raised on macOS by @dkokkotas in #159
• Context change for SFT by @xingdi-eric-yuan in #162
• Fix PDB indentation mismatch in list output context by @Copilot in #161
• Fix init obs by @xingdi-eric-yuan in #171
• Integrating thinking by @xingdi-eric-yuan in #172
• Show pytest traceback for test failures. by @MarcCote in #173
• Rich logger by @matheper in #170
• Agents rich progress by @matheper in #174
• A set of fixes by @xingdi-eric-yuan in #175
• Add get_problem_ids (formerly get_dataset_split) to all benchmark env by @MarcCote in #176
• Pin swe-smith version by @matheper in #181
• Change to when tool call is on auto parsing by @icwhite in #182
• Add memory limit to Docker containers by @matheper in #183
• Disable rich live in human mode by @matheper in #180
• Pin dataset revision for SWE-Smith dataset by @MarcCote in #188
• Fix progress skip by @matheper in #186
• Agent timeout by @matheper in #189
• Always use debug level when logging to file by @matheper in #191
• Add logfile to progress tracker. Stop pending tasks' spinner. by @MarcCote in #190
• PyPI release action by @matheper in #192
• Use hasattr instead of check dict to support subclassed methods. by @MarcCote in #194
• Fix release pipeline by @matheper in #195
• Dump experiment info by @matheper in #193
• Avoid copying pycache in the temp workding dir. by @MarcCote in #196
• Use the main logger when checking if LLM config is correct by @MarcCote in #199
• Fix get_problem_ids to respect instance_ids parameter in SWE environments by @Copilot in #179
• Improve gather_result, remove filter, fix inf loop in truncation by @xingdi-eric-yuan in #167
• Add back the message and LLM response to log by @xingdi-eric-yuan in #200
• Dump problem status by @matheper in #197
• Fix log none plus tool call by @sordonia in #202
• Fix Skip by @xingdi-eric-yuan in #211
• Refac reasoning content by @xingdi-eric-yuan in #207
• Set reset_prompt_history_after_rewrite to be False by default by @xingdi-eric-yuan in #215
• Bump version 1.1.0 by @matheper in #214

New Contributors

• @ShiZhengyan made their first contribution in #123
• @Copilot made their first contribution in #129
• @threewisemonkeys-as made their first contribution in #166
• @dkokkotas made their first contribution in #159
• @icwhite made their first contribution in #182

Full Changelog: 1.0.0...1.1.0

1.1.0rc1

What's Changed

• arxiv url by @xingdi-eric-yuan in #106
• Readme install from pypi by @matheper in #108
• Change credential order in AzureOpenAILLM for proper token retrieval by @matheper in #110
• AzureOpenAI Chained Credential by @matheper in #111
• Function Calling Syntax by @xingdi-eric-yuan in #109
• Update human class to use tool calls by @matheper in #114
• make sure the working dir is always in sys path by @xingdi-eric-yuan in #115
• Validate human input by @matheper in #116
• Bugfix human test by @matheper in #118
• Disable strict mode by @xingdi-eric-yuan in #117
• Fix env.rewrite_counter by @matheper in #119
• Fixed EvalTool being called with extra kwargs when reacting to events by @matheper in #120
• Add logs viewer for Froggy by @MarcCote in #95
• Fix: Use json.dumps for tool arguments in OpenAILLM by @ShiZhengyan in #123
• Tools observations by @matheper in #121
• Fix 'utf-8' codec error with surrogate pairs in Unicode strings by @Copilot in #129
• Add max_retries parameter to Human class to limit terminal read attempts by @Copilot in #126
• Improve test coverage report and default pytest configs by @matheper in #98
• Clean up pytest.ini, partially reverting #98 by @matheper in #132
• Remove current file by @matheper in #127
• Add start and end args to ViewTool by @matheper in #133
• Minor Fixes by @xingdi-eric-yuan in #124
• Fix kwargs by @xingdi-eric-yuan in #135
• Fix kwargs for pdb tool by @xingdi-eric-yuan in #136
• Fix type annotation: tool_call_list should be list not dict by @ShiZhengyan in #134
• Enhance Agent logging to include step number and reason for termination by @matheper in #137
• Improve visualization by @xingdi-eric-yuan in #138
• Pdb current frame file by @matheper in #139
• Pdb breakpoint handling by @matheper in #140
• Refactor llm_api into debug_gym.llms subpackage by @MarcCote in #142
• Resolve absolute path from RepoEnv by @matheper in #144
• Use better command completion for Human Mode by @MarcCote in #143
• Fix: resolve_path and is_editable to account for ignored and read-only files by @matheper in #145
• Ignore files from .gitignore by @matheper in #146
• Fix Aider ignore patterns and add tests for path resolution and ignored/read-only files by @matheper in #147
• Fix issue resolving env.working_dir by @matheper in #148
• Set default RepoEnv.dir_tree_depth to 1 by @matheper in #150
• replace unescape by filtering non-utf8 chars in system prompts by @xingdi-eric-yuan in #151
• Adding SWE-Smith support by @MarcCote in #122
• Only load image for instance_id we want to tests by @MarcCote in #154
• No eval shortcut by @matheper in #152
• Parallel execution by @matheper in #153
• Trajectory Filtering by @xingdi-eric-yuan in #141
• For SWE-Smith, add a new test split distinct from train-789 by @MarcCote in #156
• Fix ViewTool handling empty files by @matheper in #157
• Improve Retry by @xingdi-eric-yuan in #158
• Support pickling PDBTool instances by @threewisemonkeys-as in #166
• Resolve path mismatch issue raised on macOS by @dkokkotas in #159
• Context change for SFT by @xingdi-eric-yuan in #162
• Fix PDB indentation mismatch in list output context by @Copilot in #161
• Fix init obs by @xingdi-eric-yuan in #171
• Integrating thinking by @xingdi-eric-yuan in #172
• Show pytest traceback for test failures. by @MarcCote in #173
• Rich logger by @matheper in #170
• Agents rich progress by @matheper in #174
• A set of fixes by @xingdi-eric-yuan in #175
• Add get_problem_ids (formerly get_dataset_split) to all benchmark env by @MarcCote in #176
• Pin swe-smith version by @matheper in #181
• Change to when tool call is on auto parsing by @icwhite in #182
• Add memory limit to Docker containers by @matheper in #183
• Disable rich live in human mode by @matheper in #180
• Pin dataset revision for SWE-Smith dataset by @MarcCote in #188
• Fix progress skip by @matheper in #186
• Agent timeout by @matheper in #189
• Always use debug level when logging to file by @matheper in #191
• Add logfile to progress tracker. Stop pending tasks' spinner. by @MarcCote in #190
• PyPI release action by @matheper in #192
• Use hasattr instead of check dict to support subclassed methods. by @MarcCote in #194
• Fix release pipeline by @matheper in #195
• Dump experiment info by @matheper in #193
• Avoid copying pycache in the temp workding dir. by @MarcCote in #196

New Contributors

• @ShiZhengyan made their first contribution in #123
• @Copilot made their first contribution in #129
• @threewisemonkeys-as made their first contribution in #166
• @dkokkotas made their first contribution in #159
• @icwhite made their first contribution in #182

Full Changelog: 1.0.0...1.1.0

1.0.0

What's Changed

• New reasoning tool by @mormio in #1
• SubstitutionPatcher with file path by @xingdi-eric-yuan in #2
• update readme by @xingdi-eric-yuan in #3
• Tokenizer fix by @xingdi-eric-yuan in #4
• catch exceptions in max score compute by @xingdi-eric-yuan in #5
• default max score 1, in case the test crashes by @xingdi-eric-yuan in #6
• tests github actions by @matheper in #8
• Mm/froggyignore by @mormio in #9
• toolbox by @mormio in #10
• Fix Context Truncation by @xingdi-eric-yuan in #7
• Test Utils by @xingdi-eric-yuan in #12
• Unit Tests by @chisingh in #11
• Reset after Rewrite by @xingdi-eric-yuan in #14
• Comparisons between baseline(simple rewrite) and agent(zero-shot pdb) by @Kim-Minseon in #16
• zero shot agent that can access pdb after certain number of rewrite a… by @mormio in #15
• Terminal refac by @matheper in #13
• Terminal run bugfix by @matheper in #18
• pass entrypoint properly for aider by @sordonia in #19
• clean llm configs loading by @sordonia in #20
• support for threads in run by @sordonia in #21
• Test tools by @chisingh in #17
• add terminal to pdb when we re-add it by @mormio in #27
• Az login oai by @matheper in #32
• bug fix: 0 as line number by @xingdi-eric-yuan in #34
• Use logging instead of print. Add nice progress visualization. by @MarcCote in #35
• Clean Tools by @xingdi-eric-yuan in #36
• SWE-Bench integration (with caching) by @MarcCote in #28
• Refac tests by @matheper in #39
• Add .froggyignore and .froggyreadonly by @MarcCote in #40
• Add an optional depth argument to listdir. Fix tests by @MarcCote in #41
• Auth with ManagedIdentityCredential by @matheper in #43
• Set az and oai logging level to WARNING by @matheper in #45
• Mini nightmares by @xingdi-eric-yuan in #42
• Gist url by @chisingh in #46
• Moving Agents Out of Froggy by @xingdi-eric-yuan in #44
• Remove rich, put back tqdm with proper handling for logging. llm mess… by @MarcCote in #47
• Delete random by @matheper in #48
• Use the proper NOT_GIVEN value when calling openai API by @MarcCote in #49
• Add Dataclasses and RepoEnv Info refac by @matheper in #50
• Mini nightmares env. by @xingdi-eric-yuan in #54
• fix view by @xingdi-eric-yuan in #55
• Improve mini nightmare by @xingdi-eric-yuan in #56
• Improve mini nightmare by @xingdi-eric-yuan in #57
• Update terminal.py by @sordonia in #60
• Lite update tool syntax parser by @xingdi-eric-yuan in #62
• Fix PDB hanging by @MarcCote in #63
• Ranaming code. by @xingdi-eric-yuan in #64
• Remove ANSI codes when logging to file. by @MarcCote in #65
• EventHooks - Tools refac by @matheper in #58
• Skip swe-bench testing PRs by @matheper in #68
• Clean Reset Args by @xingdi-eric-yuan in #67
• froggy/agents, froggy/pond, tests/ by @sordonia in #70
• Fixed test and removed unused conftest by @matheper in #71
• Agents refac. by @sordonia in #72
• Better handling of shell session by @MarcCote in #73
• Minor fix utils by @xingdi-eric-yuan in #75
• Use session.is_running property instead of process.poll by @MarcCote in #77
• Improve README by @xingdi-eric-yuan in #76
• LLM Returns None by @xingdi-eric-yuan in #78
• strip a list by @xingdi-eric-yuan in #79
• Use LLMs with fixed temperature by @xingdi-eric-yuan in #80
• Max score by @matheper in #81
• Setup and session commands by @matheper in #82
• aider entrypoint by @sordonia in #83
• Set aider and mini_nightmare default entrypoints by @matheper in #84
• Parse think token by @xingdi-eric-yuan in #86
• Static analysis by @chisingh in #88
• remove syntax error in pytorch data because it's too easy. by @xingdi-eric-yuan in #90
• rename to debug_gym by @sordonia in #91
• master/slave -> server/client by @xingdi-eric-yuan in #93
• Minor obs change by @xingdi-eric-yuan in #94
• Fixes for SWEBench by @MarcCote in #85
• Analysis Scripts by @xingdi-eric-yuan in #87
• LLM API refac by @matheper in #97
• Rename Agents by @xingdi-eric-yuan in #100
• [WIP] README update by @xingdi-eric-yuan in #102
• RAI Statement by @xingdi-eric-yuan in #103
• Add debug-gym-llm-config-template entrypoint by @matheper in #101
• Add version by @MarcCote in #105
• skip tools when collecting task names by @xingdi-eric-yuan in #104

New Contributors

• @mormio made their first contribution in #1
• @xingdi-eric-yuan made their first contribution in #2
• @chisingh made their first contribution in #11
• @Kim-Minseon made their first contribution in #16
• @sordonia made their first contribution in #19
• @MarcCote made their first contribution in #35

Full Changelog: https://github.com/microsoft/debug-gym/commits/1.0.0