Business Workflow Map

Conversational memory retrieval and answer generation

What this workflow does

Interactive chat user asks a question, the system retrieves a similar archived conversation using embeddings and FAISS, answers with the LLM, and archives older exchanges back into the vector store.

Why this workflow matters

• Improves continuity of long-running conversations
• Preserves older exchanges as searchable embeddings
• Returns an LLM-generated response informed by recent and retrieved context

Implementation references

• GetUserQuestionNode cookbook/pocketflow-chat-memory/nodes.py · lines 6–35
• EmbedNode.post() cookbook/pocketflow-chat-memory/nodes.py · lines 124–143
• EmbedNode.prep() cookbook/pocketflow-chat-memory/nodes.py · lines 94–104
• AnswerNode.post() cookbook/pocketflow-chat-memory/nodes.py · lines 74–91
• RetrieveNode.prep() cookbook/pocketflow-chat-memory/nodes.py · lines 146–165
• RetrieveNode.exec() cookbook/pocketflow-chat-memory/nodes.py · lines 167–193
• GetUserQuestionNode.exec() cookbook/pocketflow-chat-memory/nodes.py · lines 15–24
• GetUserQuestionNode.prep() cookbook/pocketflow-chat-memory/nodes.py · lines 7–13
• create_chat_flow() cookbook/pocketflow-chat-memory/flow.py · lines 4–30
• run_chat_memory_demo() cookbook/pocketflow-chat-memory/main.py · lines 3–23
• AnswerNode.exec() cookbook/pocketflow-chat-memory/nodes.py · lines 65–72
• EmbedNode.exec() cookbook/pocketflow-chat-memory/nodes.py · lines 106–122
• AnswerNode.prep() cookbook/pocketflow-chat-memory/nodes.py · lines 38–63
• RetrieveNode.post() cookbook/pocketflow-chat-memory/nodes.py · lines 195–203
• GetUserQuestionNode.post() cookbook/pocketflow-chat-memory/nodes.py · lines 26–35
• call_llm() cookbook/pocketflow-chat-memory/utils/call_llm.py · lines 4–13
• get_embedding() cookbook/pocketflow-chat-memory/utils/get_embedding.py · lines 5–17
• create_index() cookbook/pocketflow-chat-memory/utils/vector_index.py · lines 4–5
• add_vector() cookbook/pocketflow-chat-memory/utils/vector_index.py · lines 7–15
• search_vectors() cookbook/pocketflow-chat-memory/utils/vector_index.py · lines 17–41

async recipe recommendation and approval loop

What this workflow does

recipe seeker user enters an ingredient, receives an LLM-generated recipe suggestion, and either accepts it or retries for another suggestion

Why this workflow matters

• reduces user effort in selecting a recipe
• supports iterative refinement until approval

Implementation references

• FetchRecipes.prep_async() cookbook/pocketflow-async-basic/nodes.py · lines 7–10
• GetApproval cookbook/pocketflow-async-basic/nodes.py · lines 42–63
• FetchRecipes.post_async() cookbook/pocketflow-async-basic/nodes.py · lines 17–21
• SuggestRecipe.post_async() cookbook/pocketflow-async-basic/nodes.py · lines 37–40
• GetApproval.post_async() cookbook/pocketflow-async-basic/nodes.py · lines 54–63
• create_flow() cookbook/pocketflow-async-basic/flow.py · lines 10–27
• GetApproval.exec_async() cookbook/pocketflow-async-basic/nodes.py · lines 49–52
• SuggestRecipe.exec_async() cookbook/pocketflow-async-basic/nodes.py · lines 30–35
• FetchRecipes cookbook/pocketflow-async-basic/nodes.py · lines 4–21
• SuggestRecipe.prep_async() cookbook/pocketflow-async-basic/nodes.py · lines 26–28
• GetApproval.prep_async() cookbook/pocketflow-async-basic/nodes.py · lines 45–47
• FetchRecipes.exec_async() cookbook/pocketflow-async-basic/nodes.py · lines 12–15

resume screening summary

What this workflow does

Batch map-reduce path that reads resume files, evaluates each resume against qualification criteria with an LLM, and produces a qualification summary.

Why this workflow matters

• reduces manual screening work
• produces a shortlist summary for review

Implementation references

• EvaluateResumesNode cookbook/pocketflow-map-reduce/nodes.py · lines 26–65
• ReadResumesNode.post() cookbook/pocketflow-map-reduce/nodes.py · lines 21–23
• EvaluateResumesNode.post() cookbook/pocketflow-map-reduce/nodes.py · lines 63–65
• ReduceResultsNode.post() cookbook/pocketflow-map-reduce/nodes.py · lines 93–105
• ReduceResultsNode.exec() cookbook/pocketflow-map-reduce/nodes.py · lines 74–91
• EvaluateResumesNode.exec() cookbook/pocketflow-map-reduce/nodes.py · lines 32–61
• ReadResumesNode.exec() cookbook/pocketflow-map-reduce/nodes.py · lines 9–19
• EvaluateResumesNode.prep() cookbook/pocketflow-map-reduce/nodes.py · lines 29–30
• ReduceResultsNode.prep() cookbook/pocketflow-map-reduce/nodes.py · lines 71–72

PDF text extraction and consolidation

What this workflow does

Batch PDF processing path that discovers PDFs, converts them to page images, extracts text from each page using a vision tool, and combines the results into final text.

Why this workflow matters

• turns PDFs into machine-readable text
• preserves page order in the output

Implementation references

• ProcessPDFBatchNode cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 8–52
• ProcessPDFBatchNode.prep() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 11–31
• LoadPDFNode cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 54–65
• CombineResultsNode.prep() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 96–97
• ExtractTextNode.post() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 89–91
• CombineResultsNode.post() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 110–112
• CombineResultsNode.exec() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 99–108
• CombineResultsNode cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 93–112
• ExtractTextNode.exec() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 76–87
• LoadPDFNode.post() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 63–65
• ExtractTextNode.prep() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 70–74
• LoadPDFNode.prep() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 57–58
• LoadPDFNode.exec() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 60–61
• ProcessPDFBatchNode.post() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 50–52
• create_single_pdf_flow() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 114–127
• ProcessPDFBatchNode.exec() cookbook/pocketflow-tool-pdf-vision/nodes.py · lines 33–48

retrieval-augmented question answering

What this workflow does

Offline document indexing path and online retrieval path that chunk documents, create embeddings and a FAISS index, then retrieve a relevant document and generate an LLM answer.

Why this workflow matters

• builds a searchable vector index
• returns a query-specific answer grounded in retrieved text

Implementation references

• get_offline_flow() cookbook/pocketflow-rag/flow.py · lines 4–14
• get_online_flow() cookbook/pocketflow-rag/flow.py · lines 16–26
• ChunkDocumentsNode.prep() cookbook/pocketflow-rag/nodes.py · lines 8–10
• ChunkDocumentsNode.exec() cookbook/pocketflow-rag/nodes.py · lines 12–14
• EmbedDocumentsNode.prep() cookbook/pocketflow-rag/nodes.py · lines 30–32
• EmbedDocumentsNode.exec() cookbook/pocketflow-rag/nodes.py · lines 34–36
• CreateIndexNode.prep() cookbook/pocketflow-rag/nodes.py · lines 46–48
• EmbedQueryNode.prep() cookbook/pocketflow-rag/nodes.py · lines 71–73
• RetrieveDocumentNode.prep() cookbook/pocketflow-rag/nodes.py · lines 87–89
• GenerateAnswerNode.prep() cookbook/pocketflow-rag/nodes.py · lines 120–122
• EmbedQueryNode.post() cookbook/pocketflow-rag/nodes.py · lines 81–84
• CreateIndexNode.post() cookbook/pocketflow-rag/nodes.py · lines 63–67
• EmbedQueryNode.exec() cookbook/pocketflow-rag/nodes.py · lines 75–79
• EmbedDocumentsNode.post() cookbook/pocketflow-rag/nodes.py · lines 38–43
• RetrieveDocumentNode.post() cookbook/pocketflow-rag/nodes.py · lines 112–117
• GenerateAnswerNode.post() cookbook/pocketflow-rag/nodes.py · lines 138–143
• ChunkDocumentsNode.post() cookbook/pocketflow-rag/nodes.py · lines 16–27
• CreateIndexNode cookbook/pocketflow-rag/nodes.py · lines 45–67
• RetrieveDocumentNode.exec() cookbook/pocketflow-rag/nodes.py · lines 91–110
• EmbedQueryNode cookbook/pocketflow-rag/nodes.py · lines 70–84
• GenerateAnswerNode.exec() cookbook/pocketflow-rag/nodes.py · lines 124–136

browser task automation in DOM mode / browser task automation in vision mode

What this workflow does

CLI operator launches a DOM-based browser agent that reads interactive page elements, asks an LLM for one action, and executes click/type/select/press actions until done / CLI operator launches a screenshot-based browser agent that captures images, asks an LLM for coordinates, and performs mouse/keyboard actions until done

Why this workflow matters

• automates browser interactions for a user-specified goal
• provides deterministic action logging through shared history
• supports operating pages not easily represented in the DOM
• uses raw pointer and keyboard actions to complete tasks

Implementation references

• main() cookbook/pocketflow-browser-agent/main.py · lines 15–51
• Observe.exec() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 19–39
• Decide.post() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 78–84
• Act.post() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 109–112
• Decide.exec() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 53–76
• Observe cookbook/pocketflow-browser-agent/nodes_dom.py · lines 13–44
• call_llm() cookbook/pocketflow-browser-agent/utils.py · lines 7–23
• _build_loop() cookbook/pocketflow-browser-agent/flow.py · lines 6–23
• Observe.prep() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 16–17
• Decide.prep() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 50–51
• Act.prep() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 90–91
• Observe.post() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 41–44
• Act.exec() cookbook/pocketflow-browser-agent/nodes_dom.py · lines 93–107
• Observe cookbook/pocketflow-browser-agent/nodes_vision.py · lines 14–26
• Act.exec() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 75–87
• Observe.post() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 23–26
• Act.post() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 89–92
• Decide.post() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 60–66
• Decide.exec() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 35–58
• Observe.prep() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 17–18
• Observe.exec() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 20–21
• Decide.prep() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 32–33
• Act.prep() cookbook/pocketflow-browser-agent/nodes_vision.py · lines 72–73
• create_dom_agent_flow() cookbook/pocketflow-browser-agent/flow.py · lines 26–28
• create_vision_agent_flow() cookbook/pocketflow-browser-agent/flow.py · lines 31–33

A2A task submission and answer delivery for a research agent

What this workflow does

A CLI or A2A client submits a text task to the PocketFlow A2A server, the task manager executes the research flow, and the server returns a completed task response or error.

Why this workflow matters

• Converts a user prompt into a tracked task
• Produces a final answer artifact after PocketFlow execution
• Provides structured JSON-RPC or SSE protocol responses

Implementation references

• cli() cookbook/pocketflow-a2a/a2a_client.py · lines 51–156
• A2AClient cookbook/pocketflow-a2a/common/client/client.py · lines 48–182
• PocketFlowTaskManager cookbook/pocketflow-a2a/task_manager.py · lines 22–113
• InMemoryTaskManager cookbook/pocketflow-a2a/common/server/task_manager.py · lines 78–276
• Task cookbook/pocketflow-a2a/common/types.py · lines 85–91
• AgentCard cookbook/pocketflow-a2a/common/types.py · lines 331–342
• A2AServer cookbook/pocketflow-a2a/common/server/server.py · lines 32–168
• TaskState cookbook/pocketflow-a2a/common/types.py · lines 11–18
• PocketFlowTaskManager.on_send_task() cookbook/pocketflow-a2a/task_manager.py · lines 27–92
• TaskPushNotificationConfig cookbook/pocketflow-a2a/common/types.py · lines 139–141
• main() cookbook/pocketflow-a2a/a2a_server.py · lines 26–87
• are_modalities_compatible() cookbook/pocketflow-a2a/common/server/utils.py · lines 9–20
• PocketFlowTaskManager._get_user_query() cookbook/pocketflow-a2a/task_manager.py · lines 102–113
• InMemoryTaskManager.upsert_task() cookbook/pocketflow-a2a/common/server/task_manager.py · lines 185–201
• PocketFlowTaskManager.on_send_task_subscribe() cookbook/pocketflow-a2a/task_manager.py · lines 94–100
• A2AServer._process_request() cookbook/pocketflow-a2a/common/server/server.py · lines 68–113
• search_web() cookbook/pocketflow-a2a/utils.py · lines 13–17
• call_llm() cookbook/pocketflow-a2a/utils.py · lines 5–11
• DecideAction.exec() cookbook/pocketflow-a2a/nodes.py · lines 15–64
• SendTaskResponse cookbook/pocketflow-a2a/common/types.py · lines 173–174
• SendTaskStreamingResponse cookbook/pocketflow-a2a/common/types.py · lines 182–183
• GetTaskResponse cookbook/pocketflow-a2a/common/types.py · lines 191–192
• CancelTaskResponse cookbook/pocketflow-a2a/common/types.py · lines 200–201
• SetTaskPushNotificationResponse cookbook/pocketflow-a2a/common/types.py · lines 209–210
• GetTaskPushNotificationResponse cookbook/pocketflow-a2a/common/types.py · lines 218–219
• JSONRPCMessage cookbook/pocketflow-a2a/common/types.py · lines 147–149
• JSONRPCRequest cookbook/pocketflow-a2a/common/types.py · lines 152–154
• JSONRPCResponse cookbook/pocketflow-a2a/common/types.py · lines 163–165
• SendTaskRequest cookbook/pocketflow-a2a/common/types.py · lines 168–170
• SendTaskStreamingRequest cookbook/pocketflow-a2a/common/types.py · lines 177–179
• GetTaskRequest cookbook/pocketflow-a2a/common/types.py · lines 186–188
• CancelTaskRequest cookbook/pocketflow-a2a/common/types.py · lines 195–197
• SetTaskPushNotificationRequest cookbook/pocketflow-a2a/common/types.py · lines 204–206
• GetTaskPushNotificationRequest cookbook/pocketflow-a2a/common/types.py · lines 213–215
• TaskStatusUpdateEvent cookbook/pocketflow-a2a/common/types.py · lines 94–98
• setup_test_project() cookbook/pocketflow-coding-agent/setup_test_project.py · lines 1097–1119
• ToolNode.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 99–100
• RunCommand.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 142–145
• DecideAction.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 75–80
• ReadFile.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 134–139
• ToolNode.post() cookbook/pocketflow-coding-agent/nodes.py · lines 101–107
• ListFiles.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 110–116
• GrepSearch.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 119–131
• DecideAction.post() cookbook/pocketflow-coding-agent/nodes.py · lines 81–96
• DecideAction.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 56–74

LLM-assisted coding repair workflow for a generated SQL mini-project

What this workflow does

A CLI coding agent creates a mini SQL project, then iteratively uses LLM-selected tools to inspect files, patch the workspace, and run commands until the task is done.

Why this workflow matters

• Automates local code repair and verification
• Produces edited source files and summarized result output
• Creates a reproducible test harness for SQL engine implementation

Implementation references

• main() cookbook/pocketflow-coding-agent/main.py · lines 5–23
• create_coding_agent_flow() cookbook/pocketflow-coding-agent/flow.py · lines 5–32
• DecideAction cookbook/pocketflow-coding-agent/nodes.py · lines 55–96
• call_llm() cookbook/pocketflow-a2a/utils.py · lines 5–11
• setup_test_project() cookbook/pocketflow-coding-agent/setup_test_project.py · lines 1097–1119
• DecideAction.post() cookbook/pocketflow-coding-agent/nodes.py · lines 81–96
• CompactHistory.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 42–43
• CompactHistory.post() cookbook/pocketflow-coding-agent/nodes.py · lines 52–53
• CompactHistory.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 44–51
• ToolNode cookbook/pocketflow-coding-agent/nodes.py · lines 98–107
• PatchValidate cookbook/pocketflow-coding-agent/nodes.py · lines 157–176
• DecideAction.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 56–74
• RunCommand cookbook/pocketflow-coding-agent/nodes.py · lines 141–145
• PatchRead.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 149–150
• PatchRead.post() cookbook/pocketflow-coding-agent/nodes.py · lines 154–155
• PatchApply.post() cookbook/pocketflow-coding-agent/nodes.py · lines 186–187
• PatchRead.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 151–153
• PatchValidate.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 158–160
• PatchApply.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 179–181
• PatchValidate.post() cookbook/pocketflow-coding-agent/nodes.py · lines 173–176
• PatchApply.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 182–185
• PatchValidate.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 161–172
• ToolNode.prep() cookbook/pocketflow-coding-agent/nodes.py · lines 99–100
• RunCommand.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 142–145
• DecideAction.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 75–80
• ReadFile.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 134–139
• ToolNode.post() cookbook/pocketflow-coding-agent/nodes.py · lines 101–107
• ListFiles.exec() cookbook/pocketflow-coding-agent/nodes.py · lines 110–116

LLM-assisted code generation and verification loop

What this workflow does

CLI user submits a coding problem, the system generates test cases, implements a run_code function, executes tests, and revises code until tests pass or iteration limit is reached.

Why this workflow matters

• Automates drafting and refinement of solution code
• Produces test-backed iterative improvement
• Surfaces final function implementation and test status

Implementation references

• main() cookbook/pocketflow-code-generator/main.py · lines 4–48
• ImplementFunction.prep() cookbook/pocketflow-code-generator/nodes.py · lines 57–58
• GenerateTestCases.post() cookbook/pocketflow-code-generator/nodes.py · lines 46–54
• GenerateTestCases.exec() cookbook/pocketflow-code-generator/nodes.py · lines 10–44
• ImplementFunction.exec() cookbook/pocketflow-code-generator/nodes.py · lines 60–98
• call_llm() cookbook/pocketflow-code-generator/utils/call_llm.py · lines 4–13
• RunTests.post() cookbook/pocketflow-code-generator/nodes.py · lines 136–163
• create_code_generator_flow() cookbook/pocketflow-code-generator/flow.py · lines 4–20
• ImplementFunction.post() cookbook/pocketflow-code-generator/nodes.py · lines 100–105
• Revise cookbook/pocketflow-code-generator/nodes.py · lines 165–266
• GenerateTestCases.prep() cookbook/pocketflow-code-generator/nodes.py · lines 7–8
• RunTests.prep() cookbook/pocketflow-code-generator/nodes.py · lines 108–112
• Revise.prep() cookbook/pocketflow-code-generator/nodes.py · lines 166–173
• RunTests.exec() cookbook/pocketflow-code-generator/nodes.py · lines 114–134
• Revise.post() cookbook/pocketflow-code-generator/nodes.py · lines 242–266
• Revise.exec() cookbook/pocketflow-code-generator/nodes.py · lines 175–240
• execute_python() cookbook/pocketflow-code-generator/utils/code_executor.py · lines 6–30

Natural-language to SQL execution workflow

What this workflow does

CLI-supplied database path and natural-language query are converted into schema-aware SQL, executed against SQLite, and retried through a debug loop on failure.

Why this workflow matters

• Automated schema inspection
• Query generation from natural language
• Error-driven correction loop until success or max attempts

Implementation references

• create_text_to_sql_flow() cookbook/pocketflow-text2sql/flow.py · lines 4–25
• GetSchema cookbook/pocketflow-text2sql/nodes.py · lines 7–32
• GenerateSQL.prep() cookbook/pocketflow-text2sql/nodes.py · lines 36–37
• GetSchema.exec() cookbook/pocketflow-text2sql/nodes.py · lines 11–26
• ExecuteSQL.exec() cookbook/pocketflow-text2sql/nodes.py · lines 72–99
• GenerateSQL.post() cookbook/pocketflow-text2sql/nodes.py · lines 58–65
• DebugSQL.post() cookbook/pocketflow-text2sql/nodes.py · lines 170–177
• ExecuteSQL.post() cookbook/pocketflow-text2sql/nodes.py · lines 101–133
• GenerateSQL.exec() cookbook/pocketflow-text2sql/nodes.py · lines 39–56
• DebugSQL.exec() cookbook/pocketflow-text2sql/nodes.py · lines 144–168
• GetSchema.prep() cookbook/pocketflow-text2sql/nodes.py · lines 8–9
• ExecuteSQL.prep() cookbook/pocketflow-text2sql/nodes.py · lines 69–70
• GetSchema.post() cookbook/pocketflow-text2sql/nodes.py · lines 28–32
• DebugSQL.prep() cookbook/pocketflow-text2sql/nodes.py · lines 136–142

Mini SQL query processing over local CSV tables

What this workflow does

Caller supplies SQL text; tokenizer/parser build an AST; executor reads CSV-backed tables and returns filtered/projected/joined/aggregated rows.

Why this workflow matters

• Produces query results from CSV-backed tables
• Supports filtering, joining, grouping, ordering, deduplication, and limiting

Implementation references

• TestExecuteBasicSelect.test_select_all() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 18–21
• TestExecuteBasicSelect.test_select_multiple_columns() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 29–32
• TestExecuteWhere.test_where_equals() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 42–45
• TestExecuteWhere.test_where_and() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 53–56
• TestExecuteLimit.test_limit() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 74–77
• TestExecuteLimit.test_limit_with_where() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 79–82
• TestExecuteJoin.test_inner_join() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 109–112
• TestExecuteJoin.test_left_join() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 114–117
• TestExecuteDistinct.test_distinct() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 121–124
• TestExecuteAggregatesWithoutGroupBy.test_max_salary() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 141–144
• TestExecuteAggregatesWithoutGroupBy.test_min_salary() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 146–149
• TestExecuteBasicSelect.test_select_single_column() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 23–27
• TestExecuteWhere.test_where_greater_than() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 36–40
• TestExecuteWhere.test_where_string_equals() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 47–51
• TestExecuteOrderBy.test_order_by_asc() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 60–64
• TestParseWhere.test_where_with_string() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 36–38
• TestParseWhere.test_where_and() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 40–42
• TestParseWhere.test_where_or() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 44–46
• TestParseAggregates.test_sum_aggregate() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 62–64
• TestParseGroupBy.test_group_by() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 68–70
• TestParseOrderBy.test_order_by_desc() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 84–86
• TestParseOrderBy.test_order_by_default_asc() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 88–90
• TestParseLimit.test_limit() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 94–96
• TestParseJoin.test_left_join() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 106–108
• TestParseDistinct.test_distinct() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 112–114
• TestParseDistinct.test_no_distinct() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 116–118
• TestParseSpecialConditions.test_is_not_null() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 143–145
• TestParseAggregates.test_count_star() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 50–53
• TestParseGroupBy.test_group_by_having() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 72–75
• TestParseOrderBy.test_order_by_asc() cookbook/pocketflow-coding-agent/test_project/test_parser.py · lines 79–82
• get_table() cookbook/pocketflow-coding-agent/test_project/database.py · lines 31–35
• init_db() cookbook/pocketflow-coding-agent/test_project/database.py · lines 37–45
• load_csv() cookbook/pocketflow-coding-agent/test_project/database.py · lines 8–29
• parse_select() cookbook/pocketflow-coding-agent/test_project/db.py · lines 124–159
• execute() cookbook/pocketflow-coding-agent/test_project/db.py · lines 326–346
• tokenize() cookbook/pocketflow-coding-agent/test_project/db.py · lines 37–63
• execute_join() cookbook/pocketflow-coding-agent/test_project/db.py · lines 349–373
• _aggregate() cookbook/pocketflow-coding-agent/test_project/db.py · lines 531–549
• TestExecuteGroupBy.test_group_by_avg() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 95–99
• TestExecuteGroupBy.test_group_by_having() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 101–105
• TestExecuteAggregatesWithoutGroupBy.test_count_all() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 135–139
• TestExecuteDistinct.test_distinct_vs_non_distinct() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 126–131
• TestExecuteGroupBy.test_group_by_count() cookbook/pocketflow-coding-agent/test_project/test_executor.py · lines 86–93
• parse() cookbook/pocketflow-coding-agent/test_project/db.py · lines 104–109

interactive agentic question answering with web search

What this workflow does

CLI user asks a question, the agent decides whether to search the web or answer immediately, and the workflow produces a final answer after optionally appending search results to context.

Why this workflow matters

• produces a synthesized answer for the user
• collects external search evidence into shared context

Implementation references

• main() cookbook/pocketflow-agent/main.py · lines 4–24
• SearchWeb.prep() cookbook/pocketflow-agent/nodes.py · lines 111–113
• AnswerQuestion.prep() cookbook/pocketflow-agent/nodes.py · lines 134–136
• DecideAction.exec.extract_yaml_block() cookbook/pocketflow-agent/nodes.py · lines 64–69
• SearchWeb.exec() cookbook/pocketflow-agent/nodes.py · lines 115–120
• DecideAction.prep() cookbook/pocketflow-agent/nodes.py · lines 7–14
• AnswerQuestion.post() cookbook/pocketflow-agent/nodes.py · lines 158–166
• SearchWeb.post() cookbook/pocketflow-agent/nodes.py · lines 122–131
• DecideAction.post() cookbook/pocketflow-agent/nodes.py · lines 97–108
• AnswerQuestion.exec() cookbook/pocketflow-agent/nodes.py · lines 138–156
• DecideAction.exec.parse_yaml_safely() cookbook/pocketflow-agent/nodes.py · lines 71–90
• DecideAction.exec() cookbook/pocketflow-agent/nodes.py · lines 16–95
• search_web_duckduckgo() cookbook/pocketflow-agent/utils.py · lines 14–18
• call_llm() cookbook/pocketflow-agent/utils.py · lines 6–12
• create_agent_flow() cookbook/pocketflow-agent/flow.py · lines 4–33

Generate and publish interactive flow visualizations

What this workflow does

Flow graph is converted to JSON, written to disk with an HTML D3 visualization, and optionally served and opened in a browser.

Why this workflow matters

• Creates a browsable graph artifact
• Serves the artifact locally for interactive viewing
• Can auto-open the visualization in a browser

Implementation references

• visualize_flow() cookbook/pocketflow-visualization/visualize.py · lines 904–950
• create_d3_visualization() cookbook/pocketflow-visualization/visualize.py · lines 210–829
• start_http_server() cookbook/pocketflow-visualization/visualize.py · lines 840–871
• flow_to_json() cookbook/pocketflow-visualization/visualize.py · lines 61–207
• serve_and_open_visualization() cookbook/pocketflow-visualization/visualize.py · lines 874–901

Tool-using conversational assistant with human-in-the-loop style routing

What this workflow does

A Gradio chat user asks for weather or hotel help; an LLM routes to tool calls, the tools enforce date constraints, and the system returns follow-up or result notifications while persisting session state.

Why this workflow matters

• Returns weather or booking results
• Persists action state across turns
• Produces follow-up prompts when required parameters are missing or the request is out of scope

Implementation references

• chat_fn() cookbook/pocketflow-gradio-hitl/main.py · lines 18–80
• DecideAction.exec() cookbook/pocketflow-gradio-hitl/nodes.py · lines 19–109
• load_conversation() cookbook/pocketflow-gradio-hitl/utils/conversation.py · lines 4–6
• save_conversation() cookbook/pocketflow-gradio-hitl/utils/conversation.py · lines 9–11
• DecideAction.post() cookbook/pocketflow-gradio-hitl/nodes.py · lines 111–145
• ResultNotification.post() cookbook/pocketflow-gradio-hitl/nodes.py · lines 235–241
• call_check_weather_api() cookbook/pocketflow-gradio-hitl/utils/call_mock_api.py · lines 5–18
• call_book_hotel_api() cookbook/pocketflow-gradio-hitl/utils/call_mock_api.py · lines 21–39
• create_flow() cookbook/pocketflow-gradio-hitl/flow.py · lines 12–29
• call_llm() cookbook/pocketflow-gradio-hitl/utils/call_llm.py · lines 11–17
• format_chat_history() cookbook/pocketflow-gradio-hitl/utils/format_chat_history.py · lines 1–28
• clear_fn() cookbook/pocketflow-gradio-hitl/main.py · lines 83–85
• CheckWeather.exec() cookbook/pocketflow-gradio-hitl/nodes.py · lines 156–158
• BookHotel.exec() cookbook/pocketflow-gradio-hitl/nodes.py · lines 181–183
• DecideAction.prep() cookbook/pocketflow-gradio-hitl/nodes.py · lines 14–17
• FollowUp.exec() cookbook/pocketflow-gradio-hitl/nodes.py · lines 206–210
• FollowUp.post() cookbook/pocketflow-gradio-hitl/nodes.py · lines 212–216
• ResultNotification.exec() cookbook/pocketflow-gradio-hitl/nodes.py · lines 229–233
• CheckWeather.prep() cookbook/pocketflow-gradio-hitl/nodes.py · lines 149–154
• BookHotel.prep() cookbook/pocketflow-gradio-hitl/nodes.py · lines 172–179
• FollowUp.prep() cookbook/pocketflow-gradio-hitl/nodes.py · lines 197–204
• ResultNotification.prep() cookbook/pocketflow-gradio-hitl/nodes.py · lines 220–227
• CheckWeather.post() cookbook/pocketflow-gradio-hitl/nodes.py · lines 160–168
• BookHotel.post() cookbook/pocketflow-gradio-hitl/nodes.py · lines 185–193

human-in-the-loop image generation

What this workflow does

A Streamlit UI sends a text prompt to an image generation node, then moves through review, approval, or regeneration states.

Why this workflow matters

• produces an image from text
• supports user approval before finalizing
• allows regeneration when the preview is unsatisfactory

Implementation references

• GenerateImageNode.post() cookbook/pocketflow-streamlit-fsm/nodes.py · lines 13–17
• generate_image() cookbook/pocketflow-streamlit-fsm/utils/generate_image.py · lines 5–17
• create_generation_flow() cookbook/pocketflow-streamlit-fsm/flow.py · lines 4–7
• GenerateImageNode cookbook/pocketflow-streamlit-fsm/nodes.py · lines 4–17

Human-in-the-loop task processing loop

What this workflow does

Registered web client submits a task, receives task_id and SSE status updates, then posts approved/rejected feedback to unblock the background workflow.

Why this workflow matters

• Immediate task acceptance with asynchronous processing
• Status visibility through server-sent events
• Feedback can unblock or redirect workflow execution

Implementation references

• get_index() cookbook/pocketflow-fastapi-hitl/server.py · lines 111–116
• submit_task() cookbook/pocketflow-fastapi-hitl/server.py · lines 118–160
• provide_feedback() cookbook/pocketflow-fastapi-hitl/server.py · lines 163–198
• run_flow_background() cookbook/pocketflow-fastapi-hitl/server.py · lines 39–94
• FeedbackRequest cookbook/pocketflow-fastapi-hitl/server.py · lines 104–105
• stream_status.event_generator() cookbook/pocketflow-fastapi-hitl/server.py · lines 210–239
• SubmitRequest cookbook/pocketflow-fastapi-hitl/server.py · lines 97–98
• FeedbackResponse cookbook/pocketflow-fastapi-hitl/server.py · lines 107–108
• SubmitResponse cookbook/pocketflow-fastapi-hitl/server.py · lines 100–102
• provide_feedback.report_error() cookbook/pocketflow-fastapi-hitl/server.py · lines 174–178
• stream_status() cookbook/pocketflow-fastapi-hitl/server.py · lines 202–245

voice chat conversation loop

What this workflow does

User speech is captured, transcribed, passed to an LLM, synthesized back to audio, and the conversation loops or ends.

Why this workflow matters

• hands-free conversational interaction
• continuous turn-taking until the session ends or an error occurs

Implementation references

• main() cookbook/pocketflow-voice-chat/main.py · lines 3–23
• CaptureAudioNode cookbook/pocketflow-voice-chat/nodes.py · lines 12–31
• CaptureAudioNode.post() cookbook/pocketflow-voice-chat/nodes.py · lines 21–31
• QueryLLMNode.prep() cookbook/pocketflow-voice-chat/nodes.py · lines 76–83
• QueryLLMNode.exec() cookbook/pocketflow-voice-chat/nodes.py · lines 85–92
• QueryLLMNode.post() cookbook/pocketflow-voice-chat/nodes.py · lines 94–103
• SpeechToTextNode.post() cookbook/pocketflow-voice-chat/nodes.py · lines 58–72
• TextToSpeechNode.post() cookbook/pocketflow-voice-chat/nodes.py · lines 129–148
• record_audio() cookbook/pocketflow-voice-chat/utils/audio_utils.py · lines 12–82
• CaptureAudioNode.exec() cookbook/pocketflow-voice-chat/nodes.py · lines 14–19
• speech_to_text_api() cookbook/pocketflow-voice-chat/utils/speech_to_text.py · lines 5–19
• call_llm() cookbook/pocketflow-voice-chat/utils/call_llm.py · lines 4–13
• text_to_speech_api() cookbook/pocketflow-voice-chat/utils/text_to_speech.py · lines 4–18
• create_voice_chat_flow() cookbook/pocketflow-voice-chat/flow.py · lines 4–25
• SpeechToTextNode.prep() cookbook/pocketflow-voice-chat/nodes.py · lines 35–41
• TextToSpeechNode.exec() cookbook/pocketflow-voice-chat/nodes.py · lines 120–127
• TextToSpeechNode.prep() cookbook/pocketflow-voice-chat/nodes.py · lines 107–118
• SpeechToTextNode.exec() cookbook/pocketflow-voice-chat/nodes.py · lines 43–56

Streaming chat interaction

What this workflow does

WebSocket client sends user messages, receives streamed LLM chunks, and accumulates conversation history in shared state.

Why this workflow matters

• Low-latency partial responses
• Conversation state persists across turns within a connection

Implementation references

• websocket_endpoint() cookbook/pocketflow-fastapi-websocket/main.py · lines 14–36
• StreamingChatNode.exec_async() cookbook/pocketflow-fastapi-websocket/nodes.py · lines 16–31
• StreamingChatNode.prep_async() cookbook/pocketflow-fastapi-websocket/nodes.py · lines 7–14
• StreamingChatNode cookbook/pocketflow-fastapi-websocket/nodes.py · lines 6–38
• stream_llm() cookbook/pocketflow-fastapi-websocket/utils/stream_llm.py · lines 4–16

text-to-SQL demo bootstrap and database seeding

What this workflow does

A CLI text-to-SQL demo ensures the ecommerce SQLite database exists, repopulates it when needed, and runs a query workflow over shared state.

Why this workflow matters

• ensures the demo database is ready before execution
• creates deterministic sample commerce data for query workflows

Implementation references

• populate_database() cookbook/pocketflow-text2sql/populate_db.py · lines 8–138
• run_text_to_sql() cookbook/pocketflow-text2sql/main.py · lines 6–41
• execute_sql() cookbook/pocketflow-tool-database/tools/database.py · lines 4–25

Document transformation to audio podcast

What this workflow does

A document-to-podcast pipeline extracts nuggets from documents, writes a dialogue script, and saves generated speech to an audio file.

Why this workflow matters

• Converts text documents into spoken content
• Writes generated audio to disk
• Handles MP3 or WAV output based on returned bytes

Implementation references

• main() cookbook/pocketflow-notebook-lm/main.py · lines 6–36
• AnalyzeDocs cookbook/pocketflow-notebook-lm/nodes.py · lines 11–30
• TextToSpeech.post() cookbook/pocketflow-notebook-lm/nodes.py · lines 82–98
• WriteScript.exec() cookbook/pocketflow-notebook-lm/nodes.py · lines 38–58
• AnalyzeDocs.exec() cookbook/pocketflow-notebook-lm/nodes.py · lines 16–26
• TextToSpeech.exec() cookbook/pocketflow-notebook-lm/nodes.py · lines 72–80
• create_podcast_flow() cookbook/pocketflow-notebook-lm/flow.py · lines 5–24
• AnalyzeDocs.prep() cookbook/pocketflow-notebook-lm/nodes.py · lines 12–14
• AnalyzeDocs.post() cookbook/pocketflow-notebook-lm/nodes.py · lines 28–30
• WriteScript.prep() cookbook/pocketflow-notebook-lm/nodes.py · lines 34–36
• TextToSpeech.prep() cookbook/pocketflow-notebook-lm/nodes.py · lines 68–70
• WriteScript.post() cookbook/pocketflow-notebook-lm/nodes.py · lines 60–64

documentation transformation pipeline

What this workflow does

A CLI documentation utility converts markdown docs into MDC files, combining guide/index and rewriting content before writing output files.

Why this workflow matters

• produces curated MDC documentation artifacts
• normalizes docs for downstream tooling and rule loading

Implementation references

• create_combined_guide() utils/update_pocketflow_mdc.py · lines 184–228
• convert_md_to_mdc() utils/update_pocketflow_mdc.py · lines 230–307
• generate_mdc_header() utils/update_pocketflow_mdc.py · lines 158–170
• has_substantive_content() utils/update_pocketflow_mdc.py · lines 172–182
• process_markdown_content() utils/update_pocketflow_mdc.py · lines 118–142
• generate_mdc_files() utils/update_pocketflow_mdc.py · lines 309–348

agentic retrieval augmented generation over bundled docs

What this workflow does

CLI user asks a question, the agent chooses whether to read another PocketFlow document from DOCS or answer, and the workflow builds context from documents before generating a final answer.

Why this workflow matters

• retrieves document summaries instead of searching externally
• synthesizes an answer from accumulated document context

Implementation references

• main() cookbook/pocketflow-agentic-rag/main.py · lines 4–25
• DecideAction.prep() cookbook/pocketflow-agentic-rag/nodes.py · lines 6–8
• ReadDoc.prep() cookbook/pocketflow-agentic-rag/nodes.py · lines 49–51
• ReadDoc.exec() cookbook/pocketflow-agentic-rag/nodes.py · lines 53–56
• ReadDoc.post() cookbook/pocketflow-agentic-rag/nodes.py · lines 58–62
• DecideAction.post() cookbook/pocketflow-agentic-rag/nodes.py · lines 38–45
• DecideAction.exec() cookbook/pocketflow-agentic-rag/nodes.py · lines 10–36
• Answer.post() cookbook/pocketflow-agentic-rag/nodes.py · lines 78–80
• ReadDoc cookbook/pocketflow-agentic-rag/nodes.py · lines 48–62
• create_agentic_rag_flow() cookbook/pocketflow-agentic-rag/flow.py · lines 4–28
• Answer.prep() cookbook/pocketflow-agentic-rag/nodes.py · lines 66–68
• Answer.exec() cookbook/pocketflow-agentic-rag/nodes.py · lines 70–76

recursive research and report synthesis

What this workflow does

CLI user requests a topic-specific research report, the planner generates multiple web search queries, the researcher extracts facts from search results, and the synthesizer either loops for more research or finalizes a report.

Why this workflow matters

• generates a report from multiple search-driven evidence snippets
• refines research through iterative loops when information is incomplete

Implementation references

• main() cookbook/pocketflow-deep-research/main.py · lines 4–24
• PlannerNode.prep() cookbook/pocketflow-deep-research/nodes.py · lines 8–9
• ResearcherNode.prep() cookbook/pocketflow-deep-research/nodes.py · lines 44–45
• SynthesizerNode.prep() cookbook/pocketflow-deep-research/nodes.py · lines 66–67
• PlannerNode.post() cookbook/pocketflow-deep-research/nodes.py · lines 36–38
• ResearcherNode.post() cookbook/pocketflow-deep-research/nodes.py · lines 56–60
• ResearcherNode.exec() cookbook/pocketflow-deep-research/nodes.py · lines 47–54
• SynthesizerNode.post() cookbook/pocketflow-deep-research/nodes.py · lines 102–111
• PlannerNode.exec() cookbook/pocketflow-deep-research/nodes.py · lines 11–34
• SynthesizerNode.exec() cookbook/pocketflow-deep-research/nodes.py · lines 69–100
• search_web() cookbook/pocketflow-deep-research/utils.py · lines 22–24
• call_llm() cookbook/pocketflow-deep-research/utils.py · lines 4–20
• create_deep_research_flow() cookbook/pocketflow-deep-research/flow.py · lines 4–29

diagram generation with repair loop

What this workflow does

An LLM generates Mermaid code, the compiler checks it, and the flow retries with error feedback until success or three failed attempts.

Why this workflow matters

• automates diagram drafting
• self-corrects syntax failures with feedback
• limits retry cost through a max-attempt policy

Implementation references

• main() cookbook/pocketflow-self-healing-mermaid/main.py · lines 4–45
• WriteChart.prep() cookbook/pocketflow-self-healing-mermaid/nodes.py · lines 10–12
• WriteChart.post() cookbook/pocketflow-self-healing-mermaid/nodes.py · lines 39–41
• CompileChart.post() cookbook/pocketflow-self-healing-mermaid/nodes.py · lines 87–106
• WriteChart cookbook/pocketflow-self-healing-mermaid/nodes.py · lines 7–41
• WriteChart.exec() cookbook/pocketflow-self-healing-mermaid/nodes.py · lines 14–37
• CompileChart.exec() cookbook/pocketflow-self-healing-mermaid/nodes.py · lines 50–85
• call_llm() cookbook/pocketflow-self-healing-mermaid/utils.py · lines 3–19
• CompileChart.prep() cookbook/pocketflow-self-healing-mermaid/nodes.py · lines 47–48
• create_mermaid_flow() cookbook/pocketflow-self-healing-mermaid/flow.py · lines 4–15

Background article generation with live progress updates

What this workflow does

Web user submits a topic to start a background article-generation job, then polls a streaming progress endpoint until completion.

Why this workflow matters

• Immediate job acceptance response
• Continuous progress visibility via SSE
• Completion or error feedback to the user

Implementation references

• start_job() cookbook/pocketflow-fastapi-background/main.py · lines 41–53
• run_article_workflow() cookbook/pocketflow-fastapi-background/main.py · lines 18–39
• get_progress.event_stream() cookbook/pocketflow-fastapi-background/main.py · lines 59–87
• get_progress() cookbook/pocketflow-fastapi-background/main.py · lines 55–97

Newsletter curation and formatting

What this workflow does

CLI or flow-driven newsletter pipeline searches the web for topics, selects stories, summarizes them, and formats a markdown newsletter.

Why this workflow matters

• Search results are condensed into a formatted digest
• Only the top four stories are retained
• LLM-generated blurbs provide short explanatory summaries

Implementation references

• main() cookbook/pocketflow-newsletter/main.py · lines 5–23
• CurateSources cookbook/pocketflow-newsletter/nodes.py · lines 12–27
• SummarizeStories.prep() cookbook/pocketflow-newsletter/nodes.py · lines 68–69
• FormatNewsletter.prep() cookbook/pocketflow-newsletter/nodes.py · lines 106–107
• FilterStories.post() cookbook/pocketflow-newsletter/nodes.py · lines 60–62
• FormatNewsletter.post() cookbook/pocketflow-newsletter/nodes.py · lines 116–118
• SummarizeStories.post() cookbook/pocketflow-newsletter/nodes.py · lines 97–100
• FormatNewsletter.exec() cookbook/pocketflow-newsletter/nodes.py · lines 109–114
• SummarizeStories.exec() cookbook/pocketflow-newsletter/nodes.py · lines 71–95
• FilterStories cookbook/pocketflow-newsletter/nodes.py · lines 30–62
• SummarizeStories cookbook/pocketflow-newsletter/nodes.py · lines 65–100
• search_web() cookbook/pocketflow-newsletter/utils.py · lines 22–24
• FilterStories.exec() cookbook/pocketflow-newsletter/nodes.py · lines 36–58
• create_newsletter_flow() cookbook/pocketflow-newsletter/flow.py · lines 4–26
• CurateSources.prep() cookbook/pocketflow-newsletter/nodes.py · lines 15–16
• FilterStories.prep() cookbook/pocketflow-newsletter/nodes.py · lines 33–34
• CurateSources.post() cookbook/pocketflow-newsletter/nodes.py · lines 25–27
• CurateSources.exec() cookbook/pocketflow-newsletter/nodes.py · lines 18–23

structured resume extraction

What this workflow does

A resume parser LLM path turns resume text plus target skills into validated structured YAML data.

Why this workflow matters

• produces machine-readable candidate profile
• enables downstream selection or review of extracted skills

Implementation references

• ResumeParserNode.prep() cookbook/pocketflow-structured-output/main.py · lines 7–12
• ResumeParserNode.post() cookbook/pocketflow-structured-output/main.py · lines 93–101
• ResumeParserNode.exec() cookbook/pocketflow-structured-output/main.py · lines 14–91
• call_llm() cookbook/pocketflow-structured-output/utils.py · lines 4–10
• ResumeParserNode cookbook/pocketflow-structured-output/main.py · lines 6–101