Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/8BitTacoSupreme/flowstate/llms.txt

Use this file to discover all available pages before exploring further.

flowstate memory provides three subcommands for managing FlowState’s persistent memory store. On each pipeline run, research findings, strategy assessments, interview decisions, and failure logs are automatically written to memory.db — a SQLite database using FTS5 full-text search with porter stemming. Subsequent runs automatically inject relevant prior knowledge into bridge prompts so research compounds over time.

​
Subcommands

Searches the memory store using SQLite FTS5 full-text search with BM25 ranking. Results are returned by relevance, highest first.

​
Synopsis

flowstate memory search <query> [OPTIONS]

​
Arguments

​
query
string
required
The search string. Uses porter stemming, so "streaming" matches entries containing "streams" and "streamed". Passed directly to the FTS5 MATCH expression.

​
Options

​
--kind
string
Filter results to a single memory kind. One of: research, strategy, decision, tool_run, insight.
​
--limit INTEGER
integer
default:"10"
Maximum number of results to return.
​
--root PATH
string
default:"."
Project root directory. Reads memory.db from this path.

​
Output

Results are displayed as a Rich table with the following columns:
ColumnDescription
ID12-character hex identifier for the memory entry
KindMemory kind: research, strategy, decision, tool_run, or insight
SummaryShort summary of the memory entry
ScoreBM25 relevance score (higher = more relevant)
SourceSource label (e.g., research_adapter, interview) or ---
If no results match, the command prints a dim message: No memories matching '<query>'.

​
Examples

Search all memories for Kafka-related content: 
flowstate memory search "kafka streams"
Search only strategy memories for architecture decisions: 
flowstate memory search "architecture" --kind strategy
Return up to 25 results for a broad research query: 
flowstate memory search "distributed systems" --limit 25

​
memory stats

Shows a count of stored memories broken down by kind, with a total row at the bottom.

​
Synopsis

flowstate memory stats [OPTIONS]

​
Options

​
--root PATH
string
default:"."
Project root directory. Reads memory.db from this path.

​
Output

┌─────────────┬───────┐
│ Kind        │ Count │
├─────────────┼───────┤
│ research    │    42 │
│ strategy    │     8 │
│ decision    │     3 │
│ tool_run    │    12 │
│ insight     │     0 │
│ total       │    65 │
└─────────────┴───────┘

​
Example

flowstate memory stats

​
memory clear

Deletes all entries from the memory store. Asks for confirmation unless --yes is passed.

​
Synopsis

flowstate memory clear [OPTIONS]

​
Options

​
--yes
flag
Skip the confirmation prompt and delete immediately.
​
--root PATH
string
default:"."
Project root directory. Modifies memory.db in this path.

​
Output

On success: 
Cleared 65 memories.
If confirmation is declined: 
Cancelled.

​
Examples

Clear with interactive confirmation: 
flowstate memory clear
Clear without prompting (useful in scripts): 
flowstate memory clear --yes
flowstate memory clear permanently deletes all stored knowledge accumulated across pipeline runs. This cannot be undone. Consider backing up memory.db before clearing if you want to preserve prior research findings.

​
Memory kinds

KindWhen stored
researchAfter a Research adapter step completes; artifact split by ## headings
strategyAfter a Strategy adapter step completes; artifact split by ## headings
decisionInterview answers (core_problem, ten_x_vision, architecture_pattern, research_focus)
tool_runWhen any pipeline step fails; stores the error for future reference
insightReserved for future use

Build docs developers (and LLMs) love

Get started for freeTalk to us