chore: docs

gptscript-ai · Oct 23, 2024 · e7b8cd3 · e7b8cd3
1 parent 95e0ca6
commit e7b8cd3
Show file tree

Hide file tree

Showing 16 changed files with 150 additions and 32 deletions.
diff --git a/docs/docs/03-architecture.md b/docs/docs/03-architecture.md
@@ -22,12 +22,10 @@ You can make use of it in the CLI by setting the `KNOW_SERVER_URL` environment v
 ## 3. Index Database
 
 The index database is an additional (relational) metadata database which keeps track of all datasets and ingested files and their relationships.
-It enables some extra convenience features but does not store the actual data (embeddings).
-The current implementation uses **SQLite**.
-It's fully embedded and does not require any additional setup.
+It enables some extra convenience features but does not store the actual data (content & embeddings).
+The current implementation uses **SQLite** by default, which is fully embedded and does not require any additional setup.
 
 ## 4. Vector Database
 
-The vector database is the main storage for the embeddings of the ingested documents along with some metadata (e.g. source file information).
-The current implementation uses [**chromem-go**](https://github.com/philippgille/chromem-go).
-It's fully embedded and does not require any additional setup.
+The vector database is the main storage for the content and embeddings of the ingested documents along with some metadata (e.g. source file information).
+The current implementation uses [**chromem-go**](https://github.com/philippgille/chromem-go) by default, which is fully embedded and does not require any additional setup.
diff --git a/docs/docs/06-databases.md b/docs/docs/06-databases.md
@@ -0,0 +1,30 @@
+---
+title: Index & Vector Databases
+---
+
+# Index & Vector Databases
+
+## Index Database
+
+The index database is an additional (relational) metadata database which keeps track of all datasets and ingested files and their relationships.
+It enables some extra convenience features but does not store the actual data (content & embeddings).
+The current implementation uses **SQLite** by default, which is fully embedded and does not require any additional setup.
+
+You can configure it by setting a database connection string via the `KNOW_INDEX_DSN` environment variable. 
+The following options are available:
+
+- [SQLite](https://www.sqlite.org/) (default): `KNOW_INDEX_DSN="sqlite:///home/me/mysqlite.db"`
+- [Postgres](https://www.postgresql.org/): `KNOW_INDEX_DSN="postgres://knowledge:knowledge@localhost:5432/knowledge?sslmode=disable"`
+
+
+## Vector Database
+
+The vector database is the main storage for the content and embeddings of the ingested documents along with some metadata (e.g. source file information).
+The current implementation uses [**chromem-go**](https://github.com/philippgille/chromem-go) by default, which is fully embedded and does not require any additional setup.
+
+You can configure it by setting a database connection string via the `KNOW_VECTOR_DSN` environment variable. 
+The following options are available:
+
+- [Chromem-Go](https://github.com/philippgille/chromem-go) (default): `KNOW_VECTOR_DSN="chromem:///path/to/directory"` (Note: we're using a customized fork of chromem-go, so some details may differ from the original project)
+- [PGVector](https://github.com/pgvector/pgvector): `KNOW_VECTOR_DSN="pgvector://knowledge:knowledge@localhost:5432/knowledge?sslmode=disable"`
+- [SQLite-Vec](https://github.com/asg017/sqlite-vec): `KNOW_VECTOR_DSN="sqlite-vec:///home/me/mysqlite.db"`
diff --git a/docs/docs/99-cmd/knowledge.md b/docs/docs/99-cmd/knowledge.md
@@ -20,12 +20,15 @@ knowledge [flags]
 * [knowledge askdir](knowledge_askdir.md)	 - Retrieve sources for a query from a dataset generated from a directory
 * [knowledge create-dataset](knowledge_create-dataset.md)	 - Create a new dataset
 * [knowledge delete-dataset](knowledge_delete-dataset.md)	 - Delete a dataset
+* [knowledge delete-file](knowledge_delete-file.md)	 - Delete a file from a dataset
 * [knowledge edit-dataset](knowledge_edit-dataset.md)	 - Edit an existing dataset
 * [knowledge export](knowledge_export.md)	 - Export one or more datasets as an archive (zip)
 * [knowledge get-dataset](knowledge_get-dataset.md)	 - Get a dataset
+* [knowledge get-file](knowledge_get-file.md)	 - Get a file from a dataset
 * [knowledge import](knowledge_import.md)	 - Import one or more datasets from an archive (zip) (default: all datasets)
 * [knowledge ingest](knowledge_ingest.md)	 - Ingest a file/directory into a dataset
 * [knowledge list-datasets](knowledge_list-datasets.md)	 - List existing datasets
+* [knowledge load](knowledge_load.md)	 - Load a file and transform it to markdown
 * [knowledge retrieve](knowledge_retrieve.md)	 - Retrieve sources for a query from a dataset
 * [knowledge version](knowledge_version.md)	 - 
 
diff --git a/docs/docs/99-cmd/knowledge_askdir.md b/docs/docs/99-cmd/knowledge_askdir.md
@@ -15,20 +15,24 @@ knowledge askdir [--path <path>] <query> [flags]
       --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
       --concurrency int                   Number of concurrent ingestion processes ($KNOW_INGEST_CONCURRENCY) (default 10)
   -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
-      --dsn string                        Server database connection string (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_DB_DSN)
+      --dedupe-func string                Name of the deduplication function to use ($KNOW_INGEST_DEDUPE_FUNC)
       --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
+      --err-on-unsupported-file           Error on unsupported file types ($KNOW_INGEST_ERR_ON_UNSUPPORTED_FILE)
       --flow string                       Flow name ($KNOW_FLOW)
-      --flows-file string                 Path to a YAML/JSON file containing ingestion/retrieval flows ($KNOW_FLOWS_FILE)
+      --flows-file string                 Path to a YAML/JSON file containing ingestion/retrieval flows ($KNOW_FLOWS_FILE) (default "blueprint:default")
   -h, --help                              help for askdir
       --ignore-extensions string          Comma-separated list of file extensions to ignore ($KNOW_INGEST_IGNORE_EXTENSIONS)
       --ignore-file string                Path to a .gitignore style file ($KNOW_INGEST_IGNORE_FILE)
       --include-hidden                    Include hidden files and directories ($KNOW_INGEST_INCLUDE_HIDDEN)
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
+  -w, --keyword strings                   Keywords that retrieved documents must contain ($KNOW_RETRIEVE_KEYWORDS)
       --no-create-dataset                 Do NOT create the dataset if it doesn't exist ($KNOW_INGEST_NO_CREATE_DATASET)
+      --no-prune                          Do not prune deleted files ($KNOW_ASKDIR_NO_PRUNE)
       --no-recursive                      Don't recursively ingest directories ($KNOW_NO_INGEST_RECURSIVE)
   -p, --path string                       Path to the directory to query ($KNOWLEDGE_CLIENT_ASK_DIR_PATH) (default ".")
       --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
   -k, --top-k int                         Number of sources to retrieve ($KNOWLEDGE_CLIENT_ASK_DIR_TOP_K) (default 10)
-      --vector-dbpath string              VectorDBPath to the vector database (default "$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DB_PATH)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
 ```
 
 ### SEE ALSO

diff --git a/docs/docs/99-cmd/knowledge_create-dataset.md b/docs/docs/99-cmd/knowledge_create-dataset.md
@@ -14,11 +14,11 @@ knowledge create-dataset <dataset-id> [flags]
 ```
       --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
   -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
-      --dsn string                        Server database connection string (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_DB_DSN)
       --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
   -h, --help                              help for create-dataset
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
       --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
-      --vector-dbpath string              VectorDBPath to the vector database (default "$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DB_PATH)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
 ```
 
 ### SEE ALSO

diff --git a/docs/docs/99-cmd/knowledge_delete-dataset.md b/docs/docs/99-cmd/knowledge_delete-dataset.md
@@ -14,11 +14,11 @@ knowledge delete-dataset <dataset-id> [flags]
 ```
       --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
   -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
-      --dsn string                        Server database connection string (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_DB_DSN)
       --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
   -h, --help                              help for delete-dataset
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
       --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
-      --vector-dbpath string              VectorDBPath to the vector database (default "$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DB_PATH)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
 ```
 
 ### SEE ALSO

diff --git a/docs/docs/99-cmd/knowledge_delete-file.md b/docs/docs/99-cmd/knowledge_delete-file.md
@@ -0,0 +1,28 @@
+---
+title: "knowledge delete-file"
+---
+## knowledge delete-file
+
+Delete a file from a dataset
+
+```
+knowledge delete-file <file-id|file-abs-path> [flags]
+```
+
+### Options
+
+```
+      --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
+  -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
+  -d, --dataset string                    Target Dataset ID ($KNOWLEDGE_CLIENT_DELETE_FILE_DATASET) (default "default")
+      --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
+  -h, --help                              help for delete-file
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
+      --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
+```
+
+### SEE ALSO
+
+* [knowledge](knowledge.md)	 - 
+
diff --git a/docs/docs/99-cmd/knowledge_edit-dataset.md b/docs/docs/99-cmd/knowledge_edit-dataset.md
@@ -14,14 +14,14 @@ knowledge edit-dataset <dataset-id> [flags]
 ```
       --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
   -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
-      --dsn string                        Server database connection string (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_DB_DSN)
       --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
   -h, --help                              help for edit-dataset
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
       --replace-metadata strings          replace metadata with key-value pairs (existing metadata will be removed) ($KNOWLEDGE_CLIENT_EDIT_DATASET_REPLACE_METADATA)
       --reset-metadata                    reset metadata to default (empty) ($KNOWLEDGE_CLIENT_EDIT_DATASET_RESET_METADATA)
       --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
       --update-metadata strings           update metadata key-value pairs (existing metadata will be updated/preserved) ($KNOWLEDGE_CLIENT_EDIT_DATASET_UPDATE_METADATA)
-      --vector-dbpath string              VectorDBPath to the vector database (default "$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DB_PATH)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
 ```
 
 ### SEE ALSO

diff --git a/docs/docs/99-cmd/knowledge_export.md b/docs/docs/99-cmd/knowledge_export.md
@@ -15,12 +15,12 @@ knowledge export <dataset-id> [<dataset-id>...] [flags]
   -a, --all                               Export all datasets ($KNOWLEDGE_CLIENT_EXPORT_DATASETS_ALL)
       --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
   -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
-      --dsn string                        Server database connection string (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_DB_DSN)
       --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
   -h, --help                              help for export
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
       --output string                     Output path ($KNOWLEDGE_CLIENT_EXPORT_DATASETS_OUTPUT) (default ".")
       --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
-      --vector-dbpath string              VectorDBPath to the vector database (default "$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DB_PATH)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
 ```
 
 ### SEE ALSO

diff --git a/docs/docs/99-cmd/knowledge_get-dataset.md b/docs/docs/99-cmd/knowledge_get-dataset.md
@@ -15,12 +15,12 @@ knowledge get-dataset <dataset-id> [flags]
       --archive string                    Path to the archive file ($KNOWLEDGE_CLIENT_GET_DATASET_ARCHIVE)
       --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
   -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
-      --dsn string                        Server database connection string (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_DB_DSN)
       --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
   -h, --help                              help for get-dataset
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
       --no-docs                           Do not include documents in output (way less verbose) ($KNOWLEDGE_CLIENT_GET_DATASET_NO_DOCS)
       --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
-      --vector-dbpath string              VectorDBPath to the vector database (default "$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DB_PATH)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
 ```
 
 ### SEE ALSO

diff --git a/docs/docs/99-cmd/knowledge_get-file.md b/docs/docs/99-cmd/knowledge_get-file.md
@@ -0,0 +1,28 @@
+---
+title: "knowledge get-file"
+---
+## knowledge get-file
+
+Get a file from a dataset
+
+```
+knowledge get-file <file-id|file-abs-path> [flags]
+```
+
+### Options
+
+```
+      --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
+  -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
+  -d, --dataset string                    Target Dataset ID ($KNOWLEDGE_CLIENT_GET_FILE_DATASET) (default "default")
+      --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
+  -h, --help                              help for get-file
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
+      --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
+```
+
+### SEE ALSO
+
+* [knowledge](knowledge.md)	 - 
+
diff --git a/docs/docs/99-cmd/knowledge_import.md b/docs/docs/99-cmd/knowledge_import.md
@@ -24,11 +24,11 @@ knowledge import <path> [<dataset-id>...] [flags]
 ```
       --auto-migrate string               Auto migrate database ($KNOW_DB_AUTO_MIGRATE) (default "true")
   -c, --config-file string                Path to the configuration file ($KNOW_CONFIG_FILE)
-      --dsn string                        Server database connection string (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_DB_DSN)
       --embedding-model-provider string   Embedding model provider ($KNOW_EMBEDDING_MODEL_PROVIDER) (default "openai")
   -h, --help                              help for import
+      --index-dsn string                  Index Database Connection string (relational DB) (default "sqlite://$XDG_DATA_HOME/gptscript/knowledge/knowledge.db") ($KNOW_INDEX_DSN)
       --server string                     URL of the Knowledge API Server ($KNOW_SERVER_URL)
-      --vector-dbpath string              VectorDBPath to the vector database (default "$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DB_PATH)
+      --vector-dsn string                 DSN to the vector database (default "chromem:$XDG_DATA_HOME/gptscript/knowledge/vector.db") ($KNOW_VECTOR_DSN)
 ```
 
 ### SEE ALSO