apunkt/trustgraph
1
0
Fork 0
mirror of https://github.com/trustgraph-ai/trustgraph.git synced 2026-05-02 20:03:19 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions
286 commits 80 branches 503 tags 113 MiB
Commit graph

91 commits

Author SHA1 Message Date
cybermaggedon
148092a6af
Fix/lock 0.11 version (#108) 
* - Locked 0.11 packages to 0.11 deps
- Added 'trustgraph' uber-package which installs the rest
- Added dependency to set package versions before building packages

* Bump version
 2024-10-04 22:12:39 +01:00
cybermaggedon
f081933217
Feature/subpackages (#80) 
* Renaming what will become the core package

* Tweaking to get  package build working

* Fix metering merge

* Rename to core directory

* Bump version.  Use namespace searching for packaging trustgraph-core

* Change references to trustgraph-core

* Forming embeddings-hf package

* Reference modules in core package.

* Build both packages to one container, bump version

* Update YAMLs
 2024-09-30 14:00:29 +01:00
cybermaggedon
14d79ef9f1
Streamline startup (#79) 
* Separate Prom metrics, different processors as different jobs

* Create producers before consumers, may streamline startup.

* Bump version

* Add Pulsar init command, will replace pulsar-admin invocations.

* Integrate tg-init-pulsar with YAMLs

* Update YAMLs
 2024-09-30 12:19:22 +01:00
Jack Colquitt
2f23309f05
Metering for all LLMs (#76) 
* Fixed VertexAI token counts

* Slight fix for Cohere token count

* Slight tweak to AzureAI

* Fix for prometheus for AzureAI
 2024-09-29 23:59:17 +01:00
cybermaggedon
74a14639bd
Feature/track processor state (#78) 
* Add a Prom metric to consumers & consumer/producers to track the running
state.

* New script, gets processor state using prometheus

* Bump version, add tg-processor-state to package

* Update templates
 2024-09-29 23:50:57 +01:00
cybermaggedon
efc364583b
Fix/graph rag uses wrong prompt (#77) 
* Fix queue name invocation, use correct names, not defaults

* Bump version

* Update templates
 2024-09-29 20:38:50 +01:00
Jack Colquitt
88246742f3
Features/metering all llms (#70) 
* Added Anthropic support and None logic

* Added Cohere API support

* Added support for Llamafile and OpenAI

* Added support for VertexAI

* Added AzureAI support
 2024-09-29 18:11:48 +01:00
cybermaggedon
2a49365482
Adds basic metering infrastructure (#68) 
* Basic metering module structure
* Token counting working for Bedrock
* Price calc using price list
* Added more models to pricelist
* Added Ollama token counts
----
Authored-by: JackColquitt <daniel@kalntera.ai>
 2024-09-28 20:48:20 +01:00
Jack Colquitt
9612a11581
Added basic Llamafile integration (#63) 
* Added basic Llamafile integration
* Added llamafile template support
* New templates following llamafile addition
---------
Co-authored-by: Cyber MacGeddon <cybermaggedon@gmail.com>
 2024-09-16 16:18:01 +01:00
Jack Colquitt
728ff7542a
Extraction upgrade (#61) 
* Added KG Topics
* Updated prompt-template
* Fixed prompt-generic
 2024-09-15 22:47:57 +01:00
cybermaggedon
375b213a54
Fix replace strings for JSON (#57) 
* Fix replace strings for JSON
* Add JSON markdown preamble/postamble in prompts
* Remove JSON chicanery in LLM  handling
* Bump version
 2024-09-05 20:49:32 +01:00
cybermaggedon
3445759598
Fix bedrock replace (#56) 
* Fix replace ignoring first output

* Bump version
 2024-09-05 20:00:38 +01:00
cybermaggedon
af5508ff68
Cannot access s error (#50) 
* Fix order of statements error
* Bump version
* Update templates
* Add missing script files
* Added missing DE support init
* Fixed bugs preventing definition extraction from working (#49).
 2024-09-03 22:10:48 +01:00
cybermaggedon
208c219962
Template rejig (#48) 
* document-rag / graph-rag refactor of templates

* Tweaking the docs and categories

* Clarify triple store vs RAG

* Tweak knowledge graph linkage

* Doc embedding for Qdrant

* Fix document RAG on Qdrant

* Fix templates

* Bump version

* Updated templates
 2024-09-03 00:09:15 +01:00
cybermaggedon
32b087fbf6
Switch Milvus for Qdrant in YAMLs (#43) 
* Qdrant working

* - Fix missing prompt templates
- Bump version
- Add Qdrant to packages

* Switch Milvus for Qdrant in config files
 2024-08-27 23:37:24 +01:00
cybermaggedon
e4c4774b5d
Extract rows and apply object embeddings (#42) 
* - Restructured the extract directories
- Added an extractor for 'rows' == a row of a table
- Added a row extractor prompt to prompter.
* Add row support to template prompter
* Row extraction working
* Bump version
* Emit extracted info
* Object embeddings store
* Invocation script
* Add script to package, remove cruft output
* Write rows to Cassandra
* Remove output cruft
 2024-08-27 21:55:12 +01:00
cybermaggedon
669aed0f8a
Added doc embedding support (#41) 
* document embedding writer & query
* Added test query for doc embeddings
* Bump version
* Added doc rag prompt
* Document RAG service
 2024-08-26 23:45:23 +01:00
cybermaggedon
0159e938a2
Update LLM text-completion duration metric (#40) 
* Added LLM duration metric, better buckets

* Added heatmap to dashboard to replace 95/97/99 chart

* Bump version
 2024-08-26 11:46:36 +01:00
cybermaggedon
d0e3fcf019
Catch LLM mismatches (#39) 
* Catch more upstream LLM issues

* Bump version
 2024-08-26 10:58:02 +01:00
cybermaggedon
acd60e95ec
Catch llm errors (#38) 
* Catch 'null' output from prompt for some values, presumably this is
caused by an upstream LLM error.

* Bump version
 2024-08-26 10:52:39 +01:00
cybermaggedon
cea8562ecf
Fix timeouts (#37) 
* Fix other timeout default settings

* Add storage-only YAML output

* Bump version
 2024-08-25 23:57:30 +01:00
cybermaggedon
3ca1defc88
Increase timeout (#36) 
* Increase timeout
* Bump version
 2024-08-25 20:45:04 +01:00
Jack Colquitt
980a4c5b93
Jamba and Cohere support for Bedrock (#34) 2024-08-25 20:44:00 +01:00
cybermaggedon
6edc3f0ee1
Prompt templates (#33) 
* Added prompt-template, allows definiton, relationships and kg query
to be specified in config / command-line.

* Bump version & add prompt-templates to YAMLs

* Apply to graph rag flow

* Break out different templates
 2024-08-23 23:34:16 +01:00
cybermaggedon
6d0776c7bb
Fix OpenAI reporting (#32) 
* Fix OpenAI reporting

* bump version
 2024-08-23 14:02:06 +01:00
cybermaggedon
380bddeb90
Fix LLM output reporting (#31) 
* Fix LLM output reporting

* Bump version
 2024-08-23 13:44:42 +01:00
cybermaggedon
1e92b2048a
Fix client missing exception import (#30) 
* Fix missing import, remove cruft imports

* Bump version
 2024-08-23 12:59:14 +01:00
cybermaggedon
e7c498be92
Fix neo4j (#29) 
* - Fix Neo4j memory
- Fix neo4j query

* Version to 0.7.6
 2024-08-23 12:49:42 +01:00
cybermaggedon
8372ff0eb6
Fix exception import fail (#28) 
* Fix a missing import

* Fix missing  import, bump version
 2024-08-22 23:59:26 +01:00
cybermaggedon
a2ae1d8820
Generate all YAML files (#24) 
* All templates generated, added missing file

* Up version
 2024-08-22 21:20:17 +01:00
cybermaggedon
305dda4463
Fix errors in previous update (#23) 
* Increase some limits
* Fix msg errors and update version
 2024-08-22 20:58:44 +01:00
cybermaggedon
1297cdb1d0
Improve request/response handling (#18) 
* Request/response error handling with common client

* Fixup error handling change
 2024-08-22 17:02:18 +01:00
cybermaggedon
0043b871ff
Added chunk_size metrics, and added metrics to dashboard (#16) 2024-08-22 00:20:24 +01:00
Jack Colquitt
c4bfd9fc8c
Parameters, Parsing, renaming YAMLs and Neo4j YAMLS (#15) 
* Added some params
* Parameter updates
* Fixed Neo4j issue
 2024-08-22 00:03:56 +01:00
cybermaggedon
7113d04f21
Add token chunker (#14) 2024-08-21 16:51:33 +01:00
Cyber MacGeddon
4cee446920 Fix timing issue in 2 other client fetches 2024-08-20 22:13:08 +01:00
Cyber MacGeddon
ba056b93ed Catch JSON parse errors in prompt processor 2024-08-20 20:51:32 +01:00
Cyber MacGeddon
20f983eec9 - Change flawed _client timeout logic which was causing major lags 
- Moved clients to trustgraph.clients to tidy the parent directory
- Version bump
 2024-08-20 17:54:11 +01:00
cybermaggedon
a38f530c5f
Rate limit handling (#11) 
* Added a rate limit exception
* Reduce request/response timeouts because looks like there are major issues
* Add rate limit exception catch to all consumers
* Version to 0.6.3
 2024-08-19 22:15:32 +01:00
cybermaggedon
d3e213f194
Add Neo4j support (#9) 
- Add triples-write-neo4j and triples-query-neo4j to interact with neo4j
- Add docker-compose-openai-neo4j to demo Neo4j working
 2024-08-14 09:06:33 +01:00
cybermaggedon
a3ea1301d6
Breakout store queries (#8) 
- Break out store queries, so not locked into a Milvus/Cassandra backend
- Break out prompting into a separate module, so that prompts can be tailored to other LLMs
- Jsonnet used to generate docker compose templates
- Version to 0.6.0
 2024-08-13 17:30:59 +01:00
cybermaggedon
fd547f7762
OpenAI integration (#7) 
* Preliminary OpenAI support
* Version to 0.5.9

---------

Co-authored-by: JackColquitt <daniel@kalntera.ai>
 2024-08-12 15:37:04 +01:00
JackColquitt
4584464e69 Added quick Cohere fix 2024-08-11 13:42:24 -07:00
Cyber MacGeddon
90fe017240 Bedrock LLM fix 2024-08-07 20:37:24 +01:00
Jack Colquitt
f349c8a72e
Bedrock, Cohere, and Mixed Support (#6) 
* Added Cohere support
* Tweaks to rag prompt
* Bedrock support
* Version to 0.5.7
 2024-08-07 19:27:37 +01:00
Cyber MacGeddon
103251dbf2 Make embedding & text-completion queues configurable in graph-rag 2024-08-05 22:37:11 +01:00
Cyber MacGeddon
963944f4aa Have configurable input/output queues on call from kg-relationships 2024-08-05 22:10:18 +01:00
Cyber MacGeddon
8386712574 Have configurable input/output queues on call from kg-definitions 2024-08-05 22:05:28 +01:00
Cyber MacGeddon
df5e7c42b9 Make embeddings-vectorize have arguments to configure the embeddings 
queue names
 2024-08-05 21:48:52 +01:00
Cyber MacGeddon
cd0671e777 Using persistent/non-persistent, tenants and namespaces. 2024-07-29 22:23:32 +01:00